Flexible device alignment detection apparatus and detection method thereof

Abstract

The application provides a flexible device alignment detection equipment and a detection method thereof, which comprises a jig platform, a shaping mechanism, a main detection module and a control module. The bearing plane of the jig platform is provided with a positioning reference edge to establish a workpiece coordinate system; the main detection module obtains reference images in a first state and a second state. The control module matches the two-state reference features in the workpiece coordinate system and obtains feature coordinates, solves initial plane rigid body transformation parameters; inversely maps the second-state feature coordinates to the first state to obtain a residual deformation variable vector, and reconstructs a residual deformation field; determines and normalizes an adaptive weight coefficient according to a residual modulus value, and solves a global alignment deviation of the shaping state relative to the workpiece coordinate system by using a weighted least square method. The equipment can improve the robustness and stability of alignment calculation when there is in-plane deformation.

Description

Technical Field

[0001] This invention relates to the field of electronic manufacturing, and in particular to a flexible device alignment detection device and its detection method. Background Technology

[0002] Flexible devices (such as flexible display panels and flexible circuit boards) possess structural characteristics such as thinness and bendability, and are widely used in the electronics manufacturing industry. To ensure mounting accuracy, alignment detection is often required before device processing or assembly. In existing technologies, flexible devices are typically placed on a fixture platform. A workpiece coordinate system is established using a positioning structure on the fixture platform. While the device is in its shaped state, a vision module is used to acquire image information of its upper edge reference area or alignment mark area. Subsequently, the feature coordinates of image feature points are extracted in the established coordinate system, and the planar pose of the device is calculated using image matching or template registration algorithms to guide subsequent mounting or positioning operations.

[0003] However, in practical applications, the transition of flexible devices from an unshaped state to a shaped state is often accompanied by significant non-rigid deformations, such as local warping, in-plane stretching, or compression. Existing alignment detection methods typically perform single-state analysis based solely on images from the shaped state, neglecting the geometric mapping relationship between the image features of the device in the two states. This makes it difficult to effectively identify and model feature position changes caused by deformation. Consequently, the detection system struggles to accurately distinguish between the effects of rigid body offset and local deformation, and alignment parameter calculations are easily affected by interference, especially when the device size is large or the deformation is uneven, leading to more significant errors and impacting overall mounting accuracy. Furthermore, existing alignment detection methods generally employ an equal-weight strategy in error modeling and parameter solving, failing to adjust the weights of each reference feature point according to its deformation degree. This lack of ability to suppress outliers affects the robustness of the calculation results. Simultaneously, the shaped mechanism often uses a uniform adsorption or flattening structure, failing to differentiate adjustments based on the residual deformation in different regions, and thus cannot be combined with the detection results for closed-loop optimization control.

[0004] Therefore, current alignment detection technology for flexible devices still has shortcomings in terms of polymorphic feature matching, residual deformation modeling, adaptive weight allocation, and local control capabilities. There is an urgent need to propose an alignment detection scheme that is more adaptable and ensures higher accuracy. Summary of the Invention

[0005] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, one object of this invention is to provide a flexible device alignment detection device, comprising: a fixture platform having a bearing plane for supporting the flexible device, wherein a positioning reference edge is provided on the bearing plane for establishing a workpiece coordinate system; a shaping mechanism for causing the flexible device to transition from an unshaped state to a shaped state; a main detection module for imaging multiple preset reference regions on the flexible device to acquire reference images, wherein the preset reference regions include one or both of edge reference regions and alignment mark regions; and a control module electrically connected to the shaping mechanism and the main detection module, and configured to: acquire a first-state reference image of the unshaped state and a second-state reference image of the shaped state; match corresponding reference features in the two-state reference images in the workpiece coordinate system and calculate initial planar rigid body transformation parameters, wherein the initial planar rigid body transformation parameters... The initial planar rigid body transformation parameters are used to map the feature coordinates of the first-state reference feature to the feature coordinates of the second-state reference feature. Based on the initial planar rigid body transformation parameters, the feature coordinates of the second-state reference feature are inversely transformed and mapped to the first state. The difference between the feature coordinates of the first-state reference feature and the mapped feature coordinates of the second-state reference feature is used as the residual deformation vector of each preset reference region. Based on the residual deformation vector, the residual deformation field representing the in-plane planar deformation spatial distribution of the flexible device is reconstructed. The adaptive weighting coefficients are determined and normalized according to the magnitude of the residual deformation vectors, and the global alignment deviation of the flexible device is obtained by weighted least squares calculation. The global alignment deviation is the planar rigid body transformation parameter of the flexible device relative to the workpiece coordinate system in the shaped state. The smaller the magnitude of the residual deformation vector, the larger the corresponding adaptive weighting coefficient.

[0006] In one possible implementation, the flexible device alignment detection method is applied to the aforementioned flexible device alignment detection equipment, comprising: establishing a workpiece coordinate system based on the positioning reference edge of the fixture platform; acquiring a first-state reference image in an unshaped state, the first-state reference image including multiple preset reference regions on the flexible device, the preset reference regions including one or both of edge reference regions and alignment mark regions; bringing the flexible device into a shaped state through a shaping mechanism, and acquiring a second-state reference image in the shaped state, the second-state reference image including the multiple preset reference regions; matching corresponding reference features in the two-state reference images in the workpiece coordinate system and calculating initial planar rigid body transformation parameters; based on the initial planar rigid body... The transformation parameters inversely transform the feature coordinates of the second-state reference feature to the first state, and use the difference between the feature coordinates of the first-state reference feature and the feature coordinates of the mapped second-state reference feature as the residual deformation vector of each preset reference region; based on the residual deformation vector, the residual deformation field characterizing the in-plane planar deformation spatial distribution of the flexible device is reconstructed; the adaptive weighting coefficient is determined and normalized according to the magnitude of the residual deformation vector, and the global alignment deviation of the flexible device is obtained by weighted least squares solution, wherein the global alignment deviation is the planar rigid body transformation parameter of the flexible device relative to the workpiece coordinate system in the shaped state; wherein the smaller the magnitude of the residual deformation vector, the larger the corresponding adaptive weighting coefficient.

[0007] In one possible implementation, the fixture platform is provided with a support groove for accommodating the flexible device, the bottom surface of the support groove forms the support plane, and the side of the support groove is provided with a positioning block for planar limiting of the flexible device.

[0008] In one possible implementation, the fixture platform is a multi-station structure, with each station equipped with the bearing groove and the positioning block.

[0009] In one possible implementation, the shaping mechanism includes at least one of a vacuum adsorption component and a flattening component, for applying at least one of a negative pressure adsorption force and a pressing force to the flexible device to obtain the shaped state.

[0010] In one possible implementation, the vacuum adsorption assembly includes multiple independent vacuum chambers, each of which is connected to the bearing plane through a corresponding array of vacuum adsorption holes, and each of the vacuum chambers is connected to a negative pressure source through an independent negative pressure channel to achieve zoned adjustable negative pressure adsorption.

[0011] In one possible implementation, the control module is further configured to: establish a spatial correspondence between the vacuum cavity region and a preset reference region, and determine the residual deformation index of each vacuum cavity region based on the residual deformation field. The residual deformation index is the mean of the residual deformation vector magnitude of the preset reference region corresponding to the vacuum cavity region, or the weighted mean of the residual deformation vector magnitude of the preset reference region corresponding to the vacuum cavity region. The weight of the weighted mean is an adaptive weight coefficient of the corresponding preset reference region, which is normalized within the preset reference region corresponding to the vacuum cavity region so that its sum is one. The module adjusts at least one of the negative pressure magnitude, adsorption start time, and adsorption duration of the corresponding vacuum cavity region according to the residual deformation index, and after adjustment, re-acquires the second-state reference image and updates the residual deformation field until the residual deformation index is not greater than a preset threshold.

[0012] In one possible implementation, a lateral detection module is further included, which includes at least a first lateral detection unit and a second lateral detection unit, wherein the first lateral detection unit and the second lateral detection unit are disposed at different heights along the height direction; the first lateral detection unit is used to image the upper surface or high-level reference area of ​​the flexible device, and the second lateral detection unit is used to image the edge, sidewall, or inclined surface area formed by the warping of the flexible device; the control module is configured to transform the feature coordinates of the reference features in the reference image acquired by the lateral detection module to the workpiece coordinate system, and use them together with the feature coordinates of the reference features in the reference image acquired by the main detection module for the reconstruction of the residual deformation field and the calculation of the global alignment deviation.

[0013] In one possible implementation, the reconstruction of the residual deformation field employs at least one of thin-plate spline interpolation and radial basis function interpolation; and the adaptive weighting coefficients are determined according to the rule that the magnitude of the residual deformation vector decreases monotonically, and all adaptive weighting coefficients are normalized so that their sum is one; wherein the adaptive weighting coefficients are determined in at least one of the following ways: the adaptive weighting coefficient is taken as the reciprocal of the sum of the magnitude of the residual deformation vector and a preset positive number; or the adaptive weighting coefficient is taken as an exponential decay function value that decreases as the magnitude of the residual deformation vector increases.

[0014] In one possible implementation, when the magnitude of the residual deformation vector of a preset reference region is greater than the elimination threshold, the adaptive weight coefficient corresponding to the preset reference region is set to zero and does not participate in the weighted least squares solution; and the process of updating the adaptive weight coefficient, weighted least squares solution, and updating the residual deformation vector is iterated once or multiple times. In each iteration, the feature coordinates of the second-state reference feature are inversely transformed and mapped to the first state based on the planar rigid body transformation parameters of the current iteration, and the residual deformation vector is recalculated accordingly. The planar rigid body transformation parameters of the current iteration are the initial planar rigid body transformation parameters obtained by matching the reference features corresponding to the two-state reference images in the workpiece coordinate system and solving them in the first iteration, and the planar rigid body transformation parameters corresponding to the global alignment deviation updated in subsequent iterations; until the change in the planar rigid body transformation parameters corresponding to the global alignment deviation obtained in two adjacent iterations is not greater than the preset convergence threshold or the iteration upper limit is reached.

[0015] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

[0016] This invention acquires a first-state reference image of the unshaped state and a second-state reference image of the shaped state in the same workpiece coordinate system. It matches the reference features of the two states and calculates the initial planar rigid body transformation parameters. First, the coordinates of the second-state reference features are inversely transformed to the first state. Then, the residual deformation vector is obtained from the coordinate difference between the two states, and the residual deformation field is reconstructed, allowing for a data-driven distinction between overall displacement and local in-plane deformation. Based on this, adaptive weights are determined and normalized according to the magnitude of the residual deformation (the smaller the residual, the larger the weight). Combined with weighted least squares calculation, the global alignment deviation of the flexible device relative to the workpiece coordinate system in the shaped state is solved. This reduces the pulling effect of high-deformation areas or abnormal points on the calculation results when non-rigid deformations such as local warping, stretching, or compression exist, improving the stability and consistency of the global alignment deviation and providing accurate alignment parameters for subsequent mounting / positioning. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure label:

[0019] 1. Fixture platform; 2. Bearing plane; 3. Shaping mechanism; 4. Vacuum adsorption assembly; 5. Vacuum chamber area; 6. Vacuum adsorption hole array; 7. Valve group; 8. Pressure sensor; 9. Flattening assembly; 10. Main detection module; 11. Lateral detection module; 12. Control module; 13. Flexible device.

[0020] Figure 1 This is a schematic diagram of the overall structure of the flexible device alignment detection device provided by the present invention;

[0021] Figure 2 This is a schematic diagram of the internal structure of the flexible device alignment detection device provided by the present invention;

[0022] Figure 3 Another perspective view of the overall structure of the flexible device alignment detection device provided by the present invention;

[0023] Figure 4 for Figure 3 A magnified view of part A in the middle;

[0024] Figure 5 A top view of the flexible device alignment detection device provided by the present invention;

[0025] Figure 6 The flowchart is a method for aligning flexible devices according to the present invention.

[0026] Figure 7 The flowchart shows the iterative solution process based on adaptive weights.

[0027] Figure 8 This is a schematic diagram of the inverse transformation mapping relationship of the baseline feature. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, features defined with "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0030] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0032] Please see Figure 1-8 In one possible implementation, the flexible device 13 alignment detection device includes a fixture platform 1, a shaping mechanism 3, a main detection module 10, and a control module 12. The fixture platform 1 has a bearing plane 2 to support the flexible device 13; a positioning reference edge is provided on the bearing plane 2 to determine the workpiece coordinate system. The shaping mechanism 3 applies a shaping effect to the flexible device 13, causing it to transition from an unshaped state to a shaped state. The main detection module 10 images multiple preset reference areas on the flexible device 13 to obtain reference images, wherein the preset reference areas include edge reference areas and / or alignment mark areas.

[0033] The control module 12 is electrically connected to the shaping mechanism 3 and the main detection module 10. The control module 12 acquires a first-state reference image and a second-state reference image, where the first state corresponds to the unshaped state and the second state corresponds to the shaped state. The first-state reference image can be acquired after the flexible device 13 is placed and initially positioned. The acquisition of the first and second states can be triggered based on the state determination of the shaping mechanism 3: for example, when the negative pressure reaches a set value and is maintained stably for a preset time, or when the flattening stroke is in place and the pressure meets the limit condition, the acquisition of the second-state reference image is triggered.

[0034] The positioning reference edge can be two mutually perpendicular reference edges, or it can be a straight reference edge combined with a reference point or reference hole. The control module 12 can perform edge extraction and straight line fitting based on the imaging results of the positioning reference edge to determine the coordinate axis direction and the origin position; alternatively, the control module 12 can directly read the calibration parameters of the fixture platform 1 to determine the workpiece coordinate system. After determining the workpiece coordinate system, the control module 12 performs distortion correction and calibration conversion on the reference image, converts the pixel coordinates into physical coordinates, and uniformly converts the two-state reference feature coordinates to the workpiece coordinate system.

[0035] The number of preset reference regions is at least two, and they are preferably distributed within the device surface. The edge reference region can be a contour edge segment, corner point, notch, positioning hole or positioning groove boundary, etc.; the alignment mark region can be a crosshair, ring, rectangle or coded mark, etc. The control module 12 extracts reference features from the two-state image and establishes a correspondence: the alignment mark region can extract the mark center, intersection point or circle center; the edge reference region can extract corner points or representative feature points obtained by edge fitting.

[0036] In the workpiece coordinate system, let the reference feature point corresponding to the first state be {Pᵢ}, and the reference feature point corresponding to the second state be {Qᵢ}, where i is the index number of the reference feature corresponding to the preset reference region, i=1…N, and N is the number of valid reference feature points; Pᵢ, Qᵢ, and Qᵢ′ (described later) all represent two-dimensional coordinates in the workpiece coordinate system. The control module 12 calculates the initial planar rigid body transformation parameters based on the point set corresponding to the two states. The initial planar rigid body transformation parameters are represented by rotation R(θo) and translation to=(txo,tyo), such that Qᵢ≈R(θo)·Pᵢ+to. The calculation can be performed using a least squares closed-form method (e.g., SVD / Procrustes), and can be combined with a robust matching strategy to suppress mismatched points. In this embodiment, the planar rigid body transformation includes rotation and translation, but does not include scale changes or affine shearing.

[0037] After obtaining the initial planar rigid body transformation parameters, the control module 12 inversely transforms the second-state reference characteristic coordinates to the first state, obtaining Qᵢ′=R(θo)-1·(Qi−to), and calculates the residual deformation vector dᵢ=Pᵢ−Qᵢ′. The residual deformation vector is used to characterize the in-plane residual non-rigid displacement difference after eliminating the overall rigid body motion, and ‖dᵢ‖ is the Euclidean norm of the residual deformation vector dᵢ. The control module 12 obtains the residual deformation field D(x,y)=(u(x,y),v(x,y)) based on {dᵢ}, where (x,y) are the workpiece coordinates, and u(x,y) and v(x,y) are the residual displacement components along the X-axis and Y-axis of the workpiece coordinate system at the corresponding positions, respectively; it can be obtained by thin plate spline interpolation or radial basis function interpolation, etc.

[0038] Control module 12 handles outliers based on ||dᵢ||: when ||dᵢ|| is greater than the rejection threshold, the corresponding point can be rejected or its weight can be reset to zero. Subsequently, control module 12 determines and normalizes the adaptive weight wᵢ based on ||dᵢ|| within the valid point set, ensuring the weights sum to one, and satisfying the condition that the smaller ||dᵢ|| is, the larger wᵢ is. The weight function can take the form of (||dᵢ|| + ε)⁻¹ or an exponential decay function, where ε is a preset positive number and ε > 0, used to avoid division by zero and improve numerical stability. The rejection threshold is used for outlier handling.

[0039] In the weighted least squares solution, with (θ,t) x ,tᵧ) are unknowns. Construct the objective function ∑wᵢ·‖Qᵢ−(R(θ)·Pᵢ+t)‖² and solve it to obtain the planar rigid body transformation parameters of the flexible device 13 relative to the workpiece coordinate system in the fixed state as the global alignment deviation, where R(θ) is a two-dimensional rotation matrix, corresponding to the rotation angle θ around the normal direction of the bearing plane 2, t=(t x ,tᵧ) is a two-dimensional translation vector. In this embodiment, the reference feature coordinates of the first state in the workpiece coordinate system are used as the reference pose, and the global alignment deviation is equivalent to the planar rigid body deviation of the second state relative to the reference pose. If necessary, the control module 12 can perform one or more iterative updates to the rigid body parameters after updating the weights: each iteration is based on the current (θ,t) inverse mapping of {Qᵢ} and recalculation of {dᵢ} and {wᵢ}, until the parameter changes between two adjacent iterations are less than the convergence threshold or the iteration limit is reached; wherein the convergence threshold is used to determine the iteration termination, and the iteration limit is the maximum number of iterations.

[0040] The global alignment deviation can be output as Δx, Δy, and Δθ, and a weighted residual statistic can be output as a confidence index for use in mounting / assembly compensation or process judgment. Δx and Δy are the translational deviations along the X and Y axes of the workpiece coordinate system, respectively, and Δθ is the rotational deviation about an axis perpendicular to the bearing plane 2.

[0041] In one possible implementation, the alignment detection method for the flexible device 13 is executed by the control module 12 in the alignment detection equipment in conjunction with the detection module and the shaping mechanism 3. The method includes the following steps executed by the control module 12: firstly, a workpiece coordinate system is established based on the positioning reference edge of the fixture platform 1, so that the reference features extracted in different states can be expressed and compared under a unified coordinate reference.

[0042] The positioning reference edge can be two mutually perpendicular reference edges. The control module 12 images the reference edge region and performs edge extraction and line fitting to obtain the reference edge direction, with the intersection point as the coordinate origin. The positioning reference edge can also be a straight reference edge combined with a reference hole or reference point. The coordinate axis direction is determined by fitting a straight line, and the origin position is determined by identifying the center of the reference hole or reference point. The control module 12 can combine the calibration information of the fixture platform 1 to perform distortion correction and scale conversion on the acquired image, thereby establishing a mapping relationship from pixel coordinates to physical coordinates in the workpiece coordinate system, so that the subsequently extracted two-state reference features are all represented by two-dimensional physical coordinates in the workpiece coordinate system. The mapping relationship can be determined by the calibration parameters of the fixture platform 1 and / or the camera calibration parameters.

[0043] After establishing the workpiece coordinate system, the control module 12 acquires a first-state reference image of the flexible device 13 in its unshaped state. This first-state reference image covers multiple preset reference regions on the flexible device 13, including edge reference regions and / or alignment mark regions. Subsequently, the shaping mechanism 3 applies a shaping action to the flexible device 13, bringing it into a shaped state. The control module 12 then acquires a second-state reference image in the shaped state, which also covers the aforementioned preset reference regions. The acquisition of the second-state reference image can be triggered based on the state determination of the shaping mechanism 3, for example, when the negative pressure reaches a set value and remains stable for a preset time, or when the flattening stroke is completed and the pressure meets the force limiting condition. This reduces the impact of transients during the shaping process on imaging stability.

[0044] The control module 12 extracts reference features from the two-state reference images and establishes a correspondence. At least two reference regions are preset, and they are preferably distributed within the device plane to avoid geometric degradation. For alignment mark regions, the mark center, intersection point, or circle center can be extracted as point features; for edge reference regions, corner points, notch endpoints can be extracted, or the intersection point, midpoint, or endpoint can be taken as feature points after fitting a straight line or arc to the edge segment. The control module 12 can establish an initial correspondence based on the spatial topological relationship of the preset reference regions, and filter the correspondence through template correlation, feature matching, or geometric consistency constraints to obtain a set of effective corresponding reference feature points for subsequent rigid body parameter calculation. This set of points is a one-to-one correspondence between the two states.

[0045] In the workpiece coordinate system, let the coordinates of the i-th reference feature point corresponding to the first state be Pᵢ, and the coordinates of the i-th reference feature point corresponding to the second state be Qᵢ, where i is the index number of the reference feature corresponding to the preset reference area, i=1…N, and N is the number of valid reference feature points; Pᵢ, Qᵢ, and Qᵢ′ (described later) are all two-dimensional physical coordinates in the workpiece coordinate system. Let R(θ) be a two-dimensional rotation matrix, corresponding to a rotation angle θ about the normal direction of the bearing plane 2; t=(t x ,tᵧ) is a two-dimensional translation vector; the initial parameters are θ0, t0=(t x0 Let ,tᵧ0) represent the initial planar rigid body transformation parameters. The control module 12 calculates the initial planar rigid body transformation parameters based on the corresponding point set, such that Qᵢ≈R(θ0)·Pᵢ+t0. The initial planar rigid body transformation parameters can be calculated using a closed-form method in the least squares sense, and can be combined with robust estimation to suppress mismatched points; the planar rigid body transformation only includes rotation and translation, and does not include scale changes and affine shear.

[0046] After obtaining the initial planar rigid body transformation parameters, the control module 12 performs an inverse transformation on the coordinates of the second-state reference feature points according to the initial parameters, mapping them to the first state to obtain Qᵢ′=R(θ0)⁻¹·(Qᵢ−t0), and calculates the residual deformation vector dᵢ=Pᵢ−Qᵢ′. Here, dᵢ is used to characterize the in-plane local residual displacement difference after eliminating the overall rigid body motion, and ‖dᵢ‖ is the Euclidean norm of the residual deformation vector dᵢ. The control module 12 reconstructs the residual deformation field D(x,y)=(u(x,y),v(x,y)) based on {dᵢ} to characterize the spatial distribution of the in-plane planar deformation of the flexible device 13; where (x,y) are the workpiece coordinates, and u(x,y) and v(x,y) are the residual displacement components along the X-axis and Y-axis of the workpiece coordinate system at the corresponding positions, respectively. The residual deformation field can be obtained by interpolating or fitting discrete residual vectors through thin plate spline interpolation or radial basis function interpolation.

[0047] Subsequently, control module 12 determines and normalizes the adaptive weight coefficient wᵢ based on ‖dᵢ‖. Within the set of valid points that have not been eliminated, the weights are normalized to a sum of one, and the smaller ‖dᵢ‖ is, the larger wᵢ becomes. The weight function can be wᵢ∝1 / (‖dᵢ‖+ε) or in an exponential decay form, where ε is a preset positive number and ε>0, used to avoid division by zero and improve numerical stability. When ‖dᵢ‖ exceeds the elimination threshold, the corresponding point can be eliminated or its weight can be reset to zero and not participate in subsequent calculations.

[0048] In the weighted least squares solution, with (θ,t) x Let ,tᵧ) be the unknowns. Minimize the objective function Σwᵢ·‖Qᵢ−(R(θ)·Pᵢ+t)‖² to solve for the unknowns, where t=(t xThe obtained planar rigid body transformation parameters are used as the global alignment deviation output, which is the planar rigid body transformation parameters of the flexible device 13 relative to the workpiece coordinate system in the shaped state. To clarify the reference relationship, in this embodiment, the reference feature coordinates of the first state in the workpiece coordinate system are used as the reference pose, so that the global alignment deviation is equivalent to the planar rigid body deviation of the second state relative to the reference pose. The initial planar rigid body transformation parameters are used to construct the residual deformation vector and adaptive weights. The final global alignment deviation is obtained by weighted least squares under weight constraints. The two can be the same or different. If necessary, the rigid body parameters can be iteratively updated once or multiple times after updating the weights: each iteration is based on the current parameters to inversely map {Qᵢ} and recalculate {dᵢ} and {wᵢ} until the parameter changes of two adjacent times are less than the convergence threshold or the iteration limit is reached; where the elimination threshold is used for outlier handling, the convergence threshold is used for iteration termination determination, and the iteration limit is the maximum number of iterations. The rejection threshold, convergence threshold, and iteration upper limit can be preset parameters and can be configured according to the size, tolerance, and imaging accuracy of the flexible device 13.

[0049] The global alignment deviation can be output as Δx, Δy and Δθ, and a weighted residual statistic can be output as a confidence index for mounting or assembly compensation or for process judgment; where Δx and Δy are the translational deviations along the X and Y axes of the workpiece coordinate system, respectively, and Δθ is the rotational deviation about the axis perpendicular to the bearing plane 2.

[0050] Please see Figure 1-5 In one possible implementation, the fixture platform 1 is provided with a support groove for accommodating the flexible device 13. The bottom surface of the support groove forms a support plane 2, which provides support for the flexible device 13 and forms an in-plane reference. Positioning blocks are provided on the sides of the support groove to limit the planar positioning of the flexible device 13, so that the placement position of the flexible device 13 within the support plane 2 is repeatable.

[0051] When the flexible device 13 is placed, the support groove defines its placement area. The contour of the support groove can be adapted to the shape of the flexible device 13. The main body of the device is supported by the bottom surface of the groove, and the edge area is less likely to be suspended, drooping, or swaying under the boundary constraints of the groove sidewall, thereby reducing the impact on subsequent imaging and alignment calculations. The bottom surface of the groove serves as the support plane 2, which helps to improve the consistency of device fit and provides stable geometric conditions for image acquisition and coordinate calculation.

[0052] The positioning blocks further limit the translation and rotation of the device within the bearing plane 2. The positioning blocks can be configured as two or more segments, cooperating with the sidewalls of the groove to form clear limiting boundaries. For example, the positioning blocks can be arranged on adjacent sides of the bearing groove to form an L-shaped limit, so that adjacent edges of the flexible device 13 abut against the corresponding positioning blocks, thereby constraining the translation of the device within the bearing plane 2 and suppressing rotation of the device about the normal direction of the bearing plane 2; alternatively, three-point or multi-point limiting methods can be used to ensure stable contact between the flexible device 13 and multiple positioning blocks, thereby improving repeatability. For flexible devices 13 with irregular shapes or notches, positioning holes, or positioning grooves, the positioning blocks can cooperate with the notch boundaries or hole / groove boundaries to achieve limiting, making the limiting reference clearer and reducing the risk of misplacement.

[0053] The relative positional relationship between the positioning stop and the positioning reference edge can be fixed by the calibration parameters of the fixture platform 1, so that the flexible device 13 has a more consistent initial pose in the workpiece coordinate system after each placement. As a result, the translation and rotation deviations introduced by the randomness of clamping between the two imaging before and after shaping are reduced, which is beneficial to improving the stability of the two-state reference feature matching, the initial planar rigid body transformation parameter calculation, and the global alignment deviation solution.

[0054] During the shaping process, the bearing groove and the positioning block constrain the overall sliding of the device. Especially when the shaping mechanism 3 applies negative pressure adsorption force and / or clamping force, the positioning block can suppress the overall displacement or rotation of the device, so that the difference before and after shaping is more reflected in the in-plane deformation of the device itself, rather than the pseudo displacement caused by clamping slippage, thereby improving the reliability of the residual deformation vector and residual deformation field, and improving the repeatability of the alignment deviation estimation.

[0055] The shape of the bearing groove can be rectangular, rounded rectangular, or irregular. The groove depth can be determined according to the thickness of the flexible device 13, the stroke of the shaping mechanism 3, and the allowable edge pressing amount. It can also be set as a stepped bottom surface to adapt to the support requirements of different areas. The positioning block can be a one-piece machined structure or a replaceable modular structure. Its working surface can be a straight surface, a rounded surface, or a composite surface with chamfers. An elastic layer can be set on the surface of the block to reduce edge stress concentration and the risk of scratches. The number and arrangement of the positioning blocks can be adjusted according to the shape and positioning requirements of the flexible device 13 to achieve a balance between limiting accuracy and assembly tolerance.

[0056] Please see Figure 1-5In one possible implementation, the fixture platform 1 has a multi-station structure. Each station is provided with a bearing groove for accommodating the flexible device 13 and a positioning block for planar positioning of the flexible device 13. The bottom surface of the bearing groove of each station constitutes the bearing plane 2 of the corresponding station, providing support for the flexible device 13 and forming an in-plane reference. The positioning block of each station is used to constrain the placement position and orientation of the flexible device 13 within the bearing plane 2, thereby improving the loading consistency of different stations.

[0057] The multi-station structure can be arranged linearly along the first direction, along a two-dimensional array, or along a circular path. The pitch, arrangement direction, and number of stations can be determined according to the equipment cycle time, detection field of view, and loading / unloading method. To ensure the consistency of detection results at different stations, a unified geometric reference system can be adopted to establish the reference between the positioning reference edge of each station and the workpiece coordinate system. For example, the installation position and direction of the positioning reference edge of each station relative to the reference surface of the fixture platform 1 are kept consistent, so that the workpiece coordinate system under different stations has a consistent axial definition and origin definition method, thereby facilitating the reuse of detection and calculation processes by the control module 12 between different stations.

[0058] In a multi-station structure, each station can correspond to an independent detection field of view or share a detection field of view. As an example, the main detection module 10 can perform fixed-point imaging on a single station and then switch to the next station, or use multiple cameras to image multiple stations in parallel; it can also achieve scanning imaging of the camera relative to multiple stations under the drive of the moving mechanism. The control module 12 can establish the transformation relationship between each station and the camera coordinate system based on the calibration parameters of the fixture platform 1, so that the reference feature coordinates collected by different stations can be uniformly transformed to their respective workpiece coordinate systems, thereby maintaining the consistency of two-state registration, residual deformation vector calculation, residual deformation field reconstruction, and global alignment deviation calculation.

[0059] Multi-station structures are beneficial for improving processing cycle time and equipment utilization. For example, while one station is performing shaping or waiting for stabilization, another station can simultaneously perform first-state or second-state reference image acquisition and processing, thereby reducing waiting time. For shaping processes requiring zoned adsorption or flattening, multi-station structures also facilitate the use of different shaping parameters or process strategies at different stations to adapt to the characteristics of different batches or specifications of flexible devices 13, while maintaining the uniformity of alignment detection methods.

[0060] The bearing grooves and positioning blocks at each workstation can be integrally machined or modularly replaceable to allow for quick tooling switching when changing the specifications of the flexible device 13. Positioning pins, reference holes, or reference surfaces can be set between each workstation for assembly and repeated positioning to reduce the accumulation of geometric errors between workstations. If necessary, each workstation can be geometrically calibrated separately and workstation compensation parameters can be stored in the control module 12 to further improve the consistency of the alignment deviation calculation results between multiple workstations.

[0061] Please see Figure 1-5 In one possible implementation, the shaping mechanism 3 includes at least one of a vacuum adsorption component 4 and a flattening component 9, used to apply at least one of a negative pressure adsorption force and a pressing force to the flexible device 13, so that the flexible device 13 enters a shaped state from an unshaped state. The shaping mechanism 3 enables the flexible device 13 to obtain a more stable fit on the bearing plane 2, thereby reducing the impact of warping, wrinkles, or local suspension on reference imaging and alignment calculation.

[0062] When the vacuum adsorption component 4 is used, it can be connected to a negative pressure source through an array of adsorption holes on the bearing plane 2, forming a negative pressure adsorption force between the flexible device 13 and the bearing plane 2, causing the flexible device 13 to adhere to the bearing plane 2. The adsorption holes can be distributed in a matrix, strip, or along the edge. The pore size, pore spacing, and distribution density can be determined according to the material stiffness, thickness, and easily deformable areas of the flexible device 13 to achieve a balance between the adhesion effect and local stress concentration. To avoid damage to the device surface caused by adsorption, chamfers or microporous diffusion structures can be provided around the adsorption holes. The bearing plane 2 can also be provided with a breathable membrane, microporous plate, or surface coating to improve the uniformity of adsorption.

[0063] When using the flattening assembly 9, the flattening assembly 9 may include a pressure plate, pressure roller, pressure head, or flexible pressure film, etc. Driven by the driving mechanism, the flattening assembly 9 applies a clamping force along the height direction, causing the flexible device 13 to adhere to the bearing plane 2 or the bottom surface of the bearing groove. The contact surface of the flattening assembly 9 can be a plane, a curved surface, or a composite surface with an elastic layer to adapt to the surface morphology of the flexible device 13 and reduce the risk of indentation. The clamping force can be achieved using displacement control, pressure control, or a force-position hybrid control method, and force limiting conditions and stroke completion conditions can be set to avoid overpressure leading to device damage or excessive stretching.

[0064] The vacuum adsorption component 4 and the flattening component 9 can be used individually or in combination. When used in combination, the vacuum adsorption component 4 can be used for initial bonding first, and then the flattening component 9 can be used to compensate for and flatten any locally raised areas; or the flattening component 9 can be used first to complete the initial bonding, and then the vacuum adsorption can be activated to maintain the bonding state and improve stability. The combined shaping can establish a stable shaping state in a shorter time and suppress the overall slippage of the device during the shaping process, making the difference between the state acquired in the subsequent first and second state reference image acquisitions more controllable.

[0065] The establishment of the shaping state can be used in conjunction with the state determination logic to trigger image acquisition. For example, when using the vacuum adsorption component 4, the second-state reference image acquisition can be triggered after the negative pressure reaches a set value and remains stable for a preset time; when using the flattening component 9, acquisition can be triggered after the flattening stroke is completed and the pressure meets the limit condition; when used in combination, a combined determination of "negative pressure stabilization + flattening completion" can be adopted to improve the consistency of the shaping state. The above determination parameters can be stored as configurable parameters in the control module 12 for adjustment to flexible devices 13 of different materials and sizes.

[0066] The structure of the shaping mechanism 3 can be adjusted according to the equipment space and cycle time. The vacuum adsorption component 4 can adopt a centralized negative pressure chamber or a distributed negative pressure channel structure; the flattening component 9 can adopt a single flattening unit coverage, a partitioned flattening unit coverage, or a movable flattening unit scanning method. To reduce the risk of particulate contamination or adsorption clogging, the vacuum adsorption component 4 can be equipped with a filter and a cleaning air path, and the bearing plane 2 can be equipped with a cleaning tank or a chip removal channel; the flattening component 9 can be equipped with replaceable contact surface consumables for maintenance and quick replacement.

[0067] Please see Figure 1-5 In one possible implementation, the vacuum adsorption assembly 4 includes multiple independent vacuum chamber regions 5. Each vacuum chamber region 5 is connected to the bearing plane 2 through its own vacuum adsorption hole array 6, and is connected to a negative pressure source through an independent negative pressure channel, thereby achieving zoned adjustable negative pressure adsorption.

[0068] The vacuum cavity region 5 can be divided into multiple adsorption regions along the bearing plane 2, for example, by row and column arrays, by strips, or by the functional areas of the flexible device 13. The adsorption hole arrays of each cavity region can be different. The hole diameter, hole spacing, and hole array density can be set according to the local stiffness, warping risk, and allowable contact stress of the corresponding region to achieve a balance between bonding effect and stress concentration. Relatively independent cavity regions can be configured for key reference areas to maintain a more stable bonding state during imaging and calculation.

[0069] Each vacuum chamber 5 has an independent negative pressure channel equipped with a valve assembly 7 to achieve start / stop and pressure regulation. The valve assembly 7 can use a solenoid valve with a throttling device to achieve segmented control, or a proportional valve to achieve continuous pressure regulation; if necessary, a pressure sensor 8 is configured to collect the negative pressure status so that the control module 12 can monitor and perform closed-loop control. Sealed isolation structures, such as sealing strips, sealing rings, or isolation ribs, are set between chambers to reduce crosstalk between chambers and ensure the independence of negative pressure regulation.

[0070] Different adsorption strategies can be implemented for different cavity regions. The adsorption sequence, target negative pressure, and holding time can be set according to the warpage distribution and placement of the flexible device 13. For example, negative pressure can be first applied to the edge region to press the edge together, and then negative pressure can be applied to the central region to complete the overall bonding; or a higher negative pressure can be first applied to locally warped areas for rapid bonding, and then the negative pressure in that area can be reduced to decrease stress concentration. To reduce slippage or wrinkling caused by instantaneous adsorption, a gradual pressure increase method can be used for each cavity region, allowing the negative pressure to smoothly increase to the target value within a preset time.

[0071] Vacuum chamber 5 can be formed by a multi-cavity structure below the bearing plane 2 and connected to independent negative pressure channels via a distribution plate or multi-cavity base. The negative pressure source can be a vacuum pump, vacuum generator, or centralized vacuum system; each chamber can be connected to the same negative pressure source and adjusted in sections via valve group 7, or multiple negative pressure sources can be configured to improve the adjustment range and response speed. The negative pressure channel can be equipped with a buffer chamber and a filter structure, and the pipe length and diameter can be matched to reduce the impact of pressure fluctuations on imaging stability.

[0072] The fit between the vacuum chamber 5 and the supporting groove and positioning block can be determined according to the shape of the flexible device 13, so that the area of ​​adsorption force action matches the support and limiting area, avoiding local stress concentration or edge curling at the limiting boundary. Through the above-mentioned partitioned adjustable negative pressure adsorption structure, more consistent conditions can be provided for establishing the shaping state, and a hardware basis can be provided for the control strategy of partitioned adjustment of negative pressure.

[0073] Please see Figure 1-5 In one possible implementation, the control module 12 establishes a spatial correspondence between the vacuum cavity region 5 and a preset reference region, and determines the residual deformation index of the corresponding region of each vacuum cavity region 5 based on the residual deformation field. The residual deformation field is reconstructed from the residual deformation vector, and the residual deformation index is obtained statistically based on the magnitude of the residual deformation vector within the region corresponding to the residual deformation field. It is used to characterize the degree of residual deformation in the area covered by the vacuum cavity region 5 and serves as the basis for adjusting the adsorption parameters of the vacuum cavity region 5.

[0074] The spatial correspondence between the vacuum chamber region 5 and the preset reference region can be determined through geometric mapping. The control module 12 can obtain the coverage range of each vacuum chamber region 5 in the workpiece coordinate system based on the calibration parameters of the fixture platform 1, and associate the position of the preset reference region or its corresponding reference feature point in the workpiece coordinate system with the coverage range, thereby determining the vacuum chamber region 5 to which each preset reference region belongs. For preset reference regions that span multiple vacuum chamber regions 5, their affiliation can be determined according to their feature point position, region center position, or region area ratio, or they can be included in the statistical range of multiple vacuum chamber regions 5 respectively.

[0075] The residual deformation index is taken as the mean of the residual deformation vector magnitudes of the preset reference region corresponding to the vacuum cavity region 5; or, the residual deformation index is taken as the weighted mean of the residual deformation vector magnitudes of the preset reference region corresponding to the vacuum cavity region 5. When using the weighted mean, the weights of the weighted mean are the adaptive weight coefficients of the corresponding preset reference region, and are normalized within the preset reference region corresponding to the vacuum cavity region 5 so that their sum is one. The preset reference region participating in the calculation can be limited to the valid reference region that has not been eliminated, and preset reference regions with magnitudes exceeding the elimination threshold are not included in the mean or their weights are reset to zero, so as to reduce the amplifying effect of local outliers on the index.

[0076] The control module 12 adjusts at least one of the following for the corresponding vacuum cavity 5 based on the residual deformation index: negative pressure, adsorption start time, and adsorption duration. The negative pressure can be increased or decreased by a preset step size or adjusted proportionally; the adsorption start time is used to control the adsorption sequence of each cavity; and the adsorption duration is used to control the holding time of the cavity during the shaping process. For cavities with concentrated residual deformation and high index, the negative pressure of the cavity is increased or its adsorption start time is advanced; for cavities with lower index and stable fit, the negative pressure is maintained or appropriately reduced to reduce stress concentration and avoid excessive stretching.

[0077] After adjusting the adsorption parameters of each vacuum chamber region 5, the control module 12 re-acquires the second-state reference image under the sizing state and recalculates the residual deformation vector based on the updated second-state reference image, thereby updating the residual deformation field and residual deformation index. The above process can continue to be executed according to the updated residual deformation index results until the residual deformation index of the corresponding region of each vacuum chamber region 5 is not greater than the preset threshold, or the preset upper limit of the number of cycles is reached. The preset threshold and the upper limit of the number of cycles can be set according to the material properties, dimensional tolerances and imaging accuracy of the flexible device 13 to achieve a balance between sizing quality and cycle time.

[0078] In one possible implementation, the flexible device 13 alignment detection device further includes a lateral detection module 11. The lateral detection module 11 includes at least a first lateral detection unit and a second lateral detection unit disposed at different heights along the height direction. The first lateral detection unit is used to image the upper surface or high-level reference region of the flexible device 13 to obtain a lateral reference image containing upper surface reference features; the second lateral detection unit is used to image the edges, sidewalls, or sloping areas formed by warping of the flexible device 13 to obtain a lateral reference image containing edge / sidewall reference features. By setting lateral imaging at different heights, reference regions at different heights can be effectively imaged even when the flexible device 13 exhibits warping, arching, or edge overhang, thereby improving the completeness and stability of reference feature acquisition.

[0079] Please see Figure 5-8In this embodiment, the control module 12 is configured to: extract reference features from the lateral reference image acquired by the lateral detection module 11 to obtain the pixel coordinates of the corresponding reference features; and convert the pixel coordinates into two-dimensional feature coordinates in the workpiece coordinate system based on the calibration parameters of the lateral detection unit. The calibration parameters may include the intrinsic and extrinsic parameters of the lateral detection unit and its spatial pose relationship with the fixture platform 1, used to realize the coordinate transformation from the lateral detection coordinate system to the workpiece coordinate system; if necessary, the transformation relationship can be corrected by combining the positioning reference edge or calibration reference feature of the fixture platform 1 to reduce the impact of assembly errors on the coordinate transformation accuracy.

[0080] Furthermore, the control module 12 is also configured to: transform the lateral detection module 11 to the reference feature coordinates in the workpiece coordinate system, and use them together with the feature coordinates of the reference features in the reference image acquired by the main detection module 10 as a reference feature set for subsequent calculations. Specifically, when establishing the correspondence between the two-state reference features, the reference features of the autonomous detection module 10 and the lateral detection module 11 are uniformly included in the corresponding matching; when calculating the residual deformation vector, the feature coordinates of the two-state corresponding reference features in the above-mentioned unified coordinate system are included in the same residual calculation framework; when reconstructing the residual deformation field, the residual deformation vector and its corresponding spatial position are used as interpolation or fitting nodes to obtain a more comprehensive in-plane planar deformation spatial distribution; and when solving the global alignment deviation, the reference features provided by the main detection module 10 and the lateral detection module 11 are used together for weighted least squares solution, thereby reducing the impact of missing features or poor local imaging on the global alignment deviation estimation in scenarios where the flexible device 13 has warping or edge features that are difficult to be stably acquired by the main detection module 10.

[0081] Optionally, when the extraction quality of some reference features collected by the lateral detection module 11 is poor due to occlusion, reflection or edge morphology changes, the reference features can be assigned a lower weight based on the feature matching residual, residual deformation vector magnitude or feature confidence, or excluded from the global alignment deviation calculation and residual deformation field reconstruction, so as to improve the robustness and repeatability of the fusion calculation.

[0082] Please see Figure 5-8In one possible implementation, the reconstruction of the residual deformation field employs at least one of thin-plate spline interpolation and radial basis function interpolation. The control module 12 acquires the residual deformation vector dᵢ corresponding to each preset reference region and uses its characteristic coordinates in the workpiece coordinate system as interpolation nodes. Two-dimensional vector interpolation is used to obtain a continuous residual deformation field D(x,y)=(u(x,y),v(x,y)), which characterizes the spatial distribution of in-plane deformation of the flexible device 13. Thin-plate spline interpolation can obtain a smooth displacement field by minimizing the smoothness constraint of the interpolation surface, suitable for describing overall smooth deformation; radial basis function interpolation can fit discrete residual displacements through basis function superposition, suitable for obtaining stable interpolation results when the node distribution is uneven. To suppress noise amplification, smoothing constraint parameters can be introduced into the interpolation process, and the interpolation results at the boundary of the bearing plane 2 can be constrained.

[0083] The adaptive weight coefficient wᵢ is determined according to the rule that the magnitude of the residual deformation vector ‖dᵢ‖ decreases monotonically, and is normalized within the corresponding point set of the valid reference region that has not been eliminated, so that the sum of the normalized weights is one. The weights are determined using at least one of the following methods: wᵢ is taken as 1 / (‖dᵢ‖+ε) and normalized for all wᵢ, where ε is a preset positive number and ε>0, used to avoid division by zero and improve numerical stability; or wᵢ is taken as an exponential decay function value that decreases as ‖dᵢ‖ increases, for example, wᵢ∝exp(−α‖dᵢ‖), where α is a preset decay coefficient, and normalized for all wᵢ. If necessary, upper and lower limits can be set for the weights to avoid numerical instability caused by excessively large or small weights for individual points.

[0084] When using weighted least squares to solve for global alignment error, control module 12 weights the registration residual terms of each reference feature point with wᵢ, so that regions with smaller residual deformation vector magnitudes have a higher proportion in the rigid body parameter solution, thereby reducing the impact of local abnormal deformations or mismatched points on global alignment error. The normalized weights are used to ensure the comparability of solutions under different point numbers or scales.

[0085] Please see Figure 5-8 In one possible implementation, when the magnitude of the residual deformation vector of a preset reference region, ‖dᵢ‖, is greater than the elimination threshold, the adaptive weight coefficient wᵢ corresponding to that preset reference region is set to zero, so that it does not participate in the weighted least squares solution. To improve the robustness of rigid body parameter estimation, the weight update, weighted least squares solution, and residual deformation vector update can be iterated once or multiple times.

[0086] In the workpiece coordinate system, let the coordinates of the i-th reference feature point in the first state be Pᵢ, and the coordinates of the corresponding point in the second state be Qᵢ. Let the two-dimensional rotation matrix R(θ) represent the rotation angle θ about the normal direction of the bearing plane 2, t=(t xLet (θ, t) be the translation vector, then the planar rigid body transformation parameters are denoted as (θ, t). In the first iteration, the initial planar rigid body transformation parameters (θ0, t0) are calculated based on the corresponding point set {(Pᵢ, Qᵢ)}, and the second-state coordinates are inversely transformed and mapped to the first state according to these parameters: Qᵢ′=R(θ0)⁻¹·(Qᵢ−t0). Based on this, the residual deformation vector dᵢ=Pᵢ−Qᵢ′ is calculated, and its magnitude ‖dᵢ‖ is calculated; when ‖dᵢ‖ is greater than the elimination threshold, the corresponding wᵢ is set to zero.

[0087] For baseline feature points that are not set to zero, the adaptive weight coefficients wᵢ are updated according to ‖dᵢ‖ and normalized within the effective point set. Then, a weighted least squares objective function is constructed with (θ,t) as unknowns and solved:

[0088] The planar rigid body transformation parameters obtained by solving ∑wᵢ·‖Qᵢ−(R(θ)·Pᵢ+t)‖² are used as the global alignment deviation for this iteration.

[0089] In subsequent iterations, the planar rigid body transformation parameters corresponding to the global alignment deviation obtained in the previous iteration are used as the planar rigid body transformation parameters for the current iteration. The inverse transformation mapping is re-executed on {Qᵢ}, {dᵢ} and its magnitude ‖dᵢ‖ are updated, and {wᵢ} is updated accordingly before weighted least squares solution is performed again, so that the residual deformation vector, weight allocation and planar rigid body transformation parameters are updated in a consistent manner during the iteration process.

[0090] The change in parameters between two adjacent iterations can be characterized by the norm of the translation difference and the rotation angle difference, for example, defined as a combination of |t_k−t_{k−1}| and |θ_k−θ_{k−1}|. Iteration terminates when the change is not greater than a preset convergence threshold, or when the number of iterations reaches the upper limit. Through the above-mentioned mechanism of zeroing outliers and iterative updates, the biasing effect of mismatched points and large local deformation regions on rigid body parameter estimation can be weakened, improving the stability and repeatability of the global alignment deviation calculation.

[0091] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0092] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A flexible device alignment detection device, characterized in that, include: The fixture platform has a bearing plane to support flexible devices, and a positioning reference edge is set on the bearing plane to determine the workpiece coordinate system; A shaping mechanism is used to apply a shaping effect to the flexible device, so that the flexible device enters a shaped state from an unshaped state; The main detection module is used to image multiple preset reference regions on the flexible device to obtain a reference image. The preset reference regions include edge reference regions and / or alignment mark regions. The control module is electrically connected to the shaping mechanism and the main detection module, and the control module is configured as follows: A first-state reference image and a second-state reference image are acquired respectively, wherein the first state corresponds to the amorphous state and the second state corresponds to the shaped state; The reference features in the first state reference image and the second state reference image are extracted and a correspondence is established. In the workpiece coordinate system, the initial planar rigid body transformation parameters are calculated based on the corresponding point set of the two states. The initial planar rigid body transformation parameters only include rotation and translation, and do not include scale changes and affine shear. After obtaining the initial planar rigid body transformation parameters, the second-state characteristic coordinates are inversely transformed and mapped to the first state. The residual deformation vector is determined by the difference between the first-state characteristic coordinates and the mapped second-state characteristic coordinates, and the residual deformation field is obtained based on the residual deformation vector. Within the valid point set that has not been eliminated, the adaptive weight coefficients are determined and normalized based on the magnitude of the residual deformation vector, so that the weights sum to one. The smaller the magnitude of the residual deformation vector, the larger the corresponding adaptive weight coefficient. With (θ,t) x Let ,tᵧ) be the unknowns. Minimize the objective function Σwᵢ·‖Qᵢ−(R(θ)·Pᵢ+t)‖² to solve for the unknowns, where t=(t x ,tᵧ), to obtain the planar rigid body transformation parameters of the flexible device relative to the workpiece coordinate system in the shaped state; Where θ is the rotation angle about the normal direction of the bearing plane, and t x tᵧ represents the translation along the X-axis of the workpiece coordinate system, and tᵧ represents the translation along the Y-axis of the workpiece coordinate system; P i Q represents the coordinates of the i-th reference feature point in the first state in the workpiece coordinate system; i R(θ) represents the coordinates of the reference feature point corresponding to the i-th reference feature point in the second state in the workpiece coordinate system; R(θ) represents the two-dimensional rotation matrix corresponding to the rotation angle θ about the normal direction of the bearing plane; w i This represents the adaptive weight coefficient corresponding to the i-th reference feature point; The global alignment deviation is the planar rigid body transformation parameter of the flexible device relative to the workpiece coordinate system in the shaped state, and can be output as Δx, Δy and Δθ.

2. The flexible device alignment detection equipment according to claim 1, characterized in that, The fixture platform is provided with a bearing groove for accommodating the flexible device. The bottom surface of the bearing groove forms the bearing plane, and the side of the bearing groove is provided with a positioning block for planar limiting of the flexible device.

3. The flexible device alignment detection equipment according to claim 2, characterized in that, The fixture platform has a multi-station structure, and each station is equipped with the bearing groove and the positioning block.

4. The flexible device alignment detection equipment according to any one of claims 1 to 3, characterized in that, The shaping mechanism includes at least one of a vacuum adsorption component and a flattening component, used to apply at least one of a negative pressure adsorption force and a pressing force to the flexible device to obtain the shaped state.

5. The flexible device alignment detection equipment according to claim 4, characterized in that, The vacuum adsorption assembly includes multiple independent vacuum chambers. Each vacuum chamber is connected to the bearing plane through a corresponding vacuum adsorption pore array, and each vacuum chamber is connected to a negative pressure source through an independent negative pressure channel to achieve zoned adjustable negative pressure adsorption.

6. The flexible device alignment detection equipment according to claim 5, characterized in that, The control module is further configured to: establish a spatial correspondence between the vacuum cavity region and the preset reference region, and determine the residual deformation index of the region corresponding to each vacuum cavity region based on the residual deformation field. The residual deformation index is the mean value of the residual deformation vector magnitude of the preset reference region corresponding to the vacuum cavity region, or the weighted mean value of the residual deformation vector magnitude of the preset reference region corresponding to the vacuum cavity region. The weight of the weighted mean value is an adaptive weight coefficient corresponding to the preset reference region, and is normalized within the preset reference region corresponding to the vacuum cavity region so that its sum is one. Adjust at least one of the negative pressure magnitude, adsorption start time, and adsorption duration of the corresponding vacuum cavity region according to the residual deformation index, and re-acquire the second state reference image and update the residual deformation field after adjustment until the residual deformation index is not greater than the preset threshold.

7. The flexible device alignment detection equipment according to claim 1, characterized in that, It also includes a lateral detection module, which includes at least a first lateral detection unit and a second lateral detection unit. The first lateral detection unit and the second lateral detection unit are arranged at different heights along the height direction. The first lateral detection unit is used to image the upper surface or high-level reference area of ​​the flexible device, and the second lateral detection unit is used to image the edge, sidewall, or inclined surface area formed by the warping of the flexible device. The control module is configured to transform the feature coordinates of the reference features in the reference image acquired by the lateral detection module to the workpiece coordinate system, and use them together with the feature coordinates of the reference features in the reference image acquired by the main detection module for the reconstruction of the residual deformation field and the calculation of the global alignment deviation.

8. A method for aligning and detecting flexible devices, applied to the flexible device alignment and detection equipment as described in claim 1, characterized in that, include: Establish the workpiece coordinate system based on the positioning reference edge of the fixture platform; A first-state reference image of the flexible device in an amorphous state is acquired. The first-state reference image covers multiple preset reference regions on the flexible device. The preset reference regions include edge reference regions and / or alignment mark regions. The flexible device is subjected to a shaping effect by a shaping mechanism to bring it into a shaped state, and a second-state reference image is acquired in the shaped state. The second-state reference image also covers the multiple preset reference areas. Reference features are extracted from the first state reference image and the second state reference image respectively, and a corresponding relationship is established to obtain the feature coordinates of the first state and the feature coordinates of the second state. The initial planar rigid body transformation parameters are calculated based on the corresponding point set in the workpiece coordinate system. The initial planar rigid body transformation parameters only include rotation and translation, and do not include scale changes and affine shear. The second-state feature coordinates are inversely transformed and mapped to the first state according to the initial planar rigid body transformation parameters, and the difference between the first-state feature coordinates and the mapped second-state feature coordinates is used as the residual deformation vector of each preset reference region. Reconstruct the residual deformation field based on the residual deformation vector; Within the valid point set that has not been eliminated, the adaptive weight coefficients are determined and normalized based on the magnitude of the residual deformation vector, so that the weights sum to one. The smaller the magnitude of the residual deformation vector, the larger the corresponding adaptive weight coefficient. With (θ,t) x Let ,tᵧ) be the unknowns. Minimize the objective function Σwᵢ·‖Qᵢ−(R(θ)·Pᵢ+t)‖² to solve for the unknowns, where t=(t x ,tᵧ), to obtain the planar rigid body transformation parameters of the flexible device relative to the workpiece coordinate system in the shaped state; Where θ is the rotation angle about the normal direction of the bearing plane, and t x tᵧ represents the translation along the X-axis of the workpiece coordinate system, and tᵧ represents the translation along the Y-axis of the workpiece coordinate system; Pi represents the coordinates of the i-th reference feature point in the first state in the workpiece coordinate system; Qi represents the coordinates of the reference feature point corresponding to the i-th reference feature point in the second state in the workpiece coordinate system; R(θ) represents the two-dimensional rotation matrix corresponding to the rotation angle θ around the normal direction of the bearing plane; wi represents the adaptive weight coefficient corresponding to the i-th reference feature point. The global alignment deviation is the planar rigid body transformation parameter of the flexible device relative to the workpiece coordinate system in the shaped state, and can be output as Δx, Δy and Δθ.

9. The method for aligning flexible devices according to claim 8, characterized in that, The reconstruction of the residual deformation field employs at least one of thin-plate spline interpolation and radial basis function interpolation. The adaptive weighting coefficients are determined according to the rule that the magnitude of the residual deformation vector decreases monotonically, and all adaptive weighting coefficients are normalized to a sum of one. The adaptive weighting coefficients are determined in at least one of the following ways: the adaptive weighting coefficient is the reciprocal of the sum of the magnitude of the residual deformation vector and a preset positive number; the adaptive weighting coefficient is an exponential decay function value that decreases as the magnitude of the residual deformation vector increases.

10. The method for aligning flexible devices according to claim 8 or 9, characterized in that, When the magnitude of the residual deformation vector of a certain preset reference region is greater than the elimination threshold, the adaptive weight coefficient corresponding to the preset reference region is set to zero and does not participate in the weighted least squares solution. The process of updating the adaptive weight coefficient, weighted least squares solution, and updating the residual deformation vector is iterated once or multiple times. In each iteration, the feature coordinates of the second-state reference feature are inversely transformed and mapped to the first state based on the planar rigid body transformation parameters of the current iteration, and the residual deformation vector is recalculated accordingly. The planar rigid body transformation parameters of the current iteration are the initial planar rigid body transformation parameters obtained by matching the corresponding reference features in the two-state reference images in the workpiece coordinate system and solving them in the first iteration. In subsequent iterations, they are the planar rigid body transformation parameters corresponding to the global alignment deviation that are iteratively updated. This continues until the change in the planar rigid body transformation parameters corresponding to the global alignment deviation obtained in two adjacent iterations is not greater than the preset convergence threshold or the iteration upper limit is reached.

Images