A method for measuring 360-degree multi-bino collaborative displacement field on column surface in underwater environment of pressure cylinder

Abstract

A method for measuring the 360° multi-binocular collaborative displacement field of a cylinder surface in an underwater environment of a pressure vessel is proposed. A composite speckle marking field is constructed on the surface of the cylinder under test. A ring-shaped marker is used to locate the center of the marker points, and random speckles within the ring are used simultaneously for identification coding and DIC matching, eliminating the need for additional marker points or calibration objects. The binocular subsystems are arranged vertically with the baseline parallel to the cylinder axis. The circumferential coverage angle and the number of systems are conservatively calculated using the shortest distance from the midpoint of the baseline to the cylinder surface. Data is acquired synchronously using a unified PTP clock and converged at an underwater node. Global coordinates of the coded marker points are obtained using photogrammetry. An initial pose is established using corresponding marker points, and the pose is optimized using the cylinder's geometric constraints. Based on the air-transparent observation window-water medium refraction light path, the corresponding rays of the upper and lower cameras are traced, and local 3D points are reconstructed through water-side ray intersection. After global coordinate system transformation, displacement calculation, and data fusion, a 360° three-dimensional displacement field on the cylinder surface is obtained.

Description

Technical Field

[0001] This invention relates to the fields of experimental mechanics, digital image correlation, underwater visual measurement, and multi-camera collaborative measurement, and in particular to a method for measuring the 360° multi-binocular collaborative displacement field of a cylinder surface in an underwater environment. Background Technology

[0002] The deformation of the cylindrical pressure-resistant structure inside the pressure cylinder under load can be measured using single-point methods such as displacement gauges, strain gauges, or laser displacement sensors. These methods offer high single-point accuracy and fast time response, but can only obtain discrete information at a limited number of locations. Increasing the spatial measurement density requires adding a large number of sensors, leads, and acquisition channels, and may still miss local deformations with uncertain locations. Full-field visual measurement methods, such as digital image correlation, can obtain continuous and dense surface displacement information non-contactly, with less additional impact on the measured structure, making them more suitable for characterizing local deformation concentrations and their evolution.

[0003] Due to the curved surface obstruction of the cylindrical structure, a single binocular system can only measure the locally visible cylindrical surface. Obtaining a 360° full-circumference displacement field requires multiple binocular subsystems arranged along the circumference of the cylinder. Given the limited space inside the pressure cylinder, if two cameras in the same group are arranged horizontally along the cylindrical surface, the circumferential observation azimuth difference between the two cameras is significant. Affected by the curvature of the cylindrical surface and self-occlusion, their shared effective field of view is small. To balance the shared field of view and imaging resolution, it is usually necessary to reduce the binocular baseline, increase the working distance, or use a longer focal length lens. The former reduces depth measurement capability, while the latter two are limited by internal space. Simultaneously, a horizontal arrangement causes the corresponding search direction to extend primarily along the circumference of the cylindrical surface, where curvature leads to significant depth variations, which is detrimental to speckle stereo matching. Therefore, it is necessary to determine a binocular arrangement more suitable for cylindrical surface measurement and to rationally determine the number of full-circumference binocular subsystems based on actual imaging conditions.

[0004] These types of tests typically involve mounting multiple cameras, lighting components, and data equipment on a measuring fixture, placing the cylindrical structure under test in the center of the fixture, and then assembling the entire unit for underwater operation. After assembly, the space between the cameras and the structure under test is limited, making it inconvenient to reposition, adjust, or remove dedicated calibration materials, or to temporarily affix markers. If additional coded markers are placed outside the speckle field, they will occupy the effective measurement area and affect DIC matching. Therefore, it is necessary to utilize the composite speckle pattern of the surface under test itself to simultaneously complete speckle measurement, marker location, and identification.

[0005] Multiple binocular subsystems establish their own local coordinate systems, determining relative poses solely based on tooling dimensions or the nominal camera mounting position. This approach is susceptible to errors in machining, assembly, and camera reconstruction, leading to misalignment of 3D data in adjacent measurement areas. Therefore, it is necessary to establish initial pose relationships using pre-measured global coordinate markers, and further utilize the geometric features of the cylindrical structure itself to correct the rotational and translational relationships between each local coordinate system and the global coordinate system.

[0006] Multi-camera systems also face challenges in camera encapsulation, synchronous acquisition, and transmission through the chamber. If each camera requires its own data interface, synchronization line, and transmission connector, it increases the number of internal cables, trigger distributors, relay devices, and external receiving interfaces, and also raises the difficulty of wiring and interface sealing. Delays in different trigger lines and relay components can also cause inconsistencies in camera exposure times, leading to the structural response at different moments during loading being mistaken for spatial displacement differences.

[0007] Using a unified clock allows all cameras to share the same time reference and expose independently at predetermined times. This reduces the need for dedicated trigger lines and relay layers, and facilitates the aggregation and synchronization of images based on acquisition sequence number and timestamp. Therefore, it is also necessary to integrate the unified clock, multi-channel buffering, synchronization frame aggregation, and single-path out-of-cabin transmission into a single design.

[0008] Furthermore, the camera's imaging light rays need to pass sequentially through air, a transparent viewing window, and water, refracting at the interfaces of these different media. Refraction errors are related to the image point position, the angle of incidence of the light rays, the thickness and normal of the viewing window, the position of the camera's optical center, and the refractive index of the medium, and are more pronounced at image edges and in tilted viewing areas. If a linear projection model based on an air environment is still used, systematic 3D reconstruction errors will occur, leading to inconsistencies in the 3D data of overlapping areas between adjacent binocular subsystems.

[0009] Therefore, a 360° multi-binocular collaborative measurement method for the column surface suitable for underwater environments of pressure cylinders is needed. This method combines binocular vertical arrangement and quantity calculation, circular composite speckle field, unified clock and data aggregation, global coordinate system, and multi-medium refraction three-dimensional reconstruction to achieve continuous, synchronous, and coordinate-unified full-circumference three-dimensional displacement measurement of the column surface. Summary of the Invention

[0010] To address the shortcomings of existing production technologies, this applicant provides a method for measuring the 360° multi-binocular collaborative displacement field of a cylinder surface in an underwater environment, effectively solving the problems of limited internal space and the difficulty in rationally arranging, synchronously acquiring, uniformly transmitting, and accurately stitching multiple binocular subsystems under conditions where the measuring fixture and the structure under test are submerged as a whole. Based on camera imaging parameters, working distance, medium refractive index, cylinder radius, and circumferential overlap requirements, this invention conservatively calculates the effective circumferential coverage of a single binocular subsystem and determines the number of systems required to achieve continuous 360° coverage. By making the baselines of the two cameras in the same group parallel or approximately parallel to the cylinder axis, the common circumferential field of view is expanded, reducing the influence of cylinder curvature on stereo matching. This invention sets up a ring on the random speckle field, uses the edge of the ring to locate the center of the marker point, and uses the random speckle inside the ring to identify the region, allowing this area to continue participating in DIC matching. This adapts to measurement conditions where it is inconvenient to set additional marker points or dedicated calibration objects after overall installation. This invention utilizes coded ring points to establish the initial pose relationship between the local and global coordinate systems of each binocular camera, and optimizes it by combining the common cylindrical geometric features of the reference state cylinder surface. Simultaneously, a unified clock is used to organize simultaneous exposures of multiple cameras, and multiple images are multiplexed into a single data stream exiting the chamber via an in-tube data aggregation node. Furthermore, this invention performs ray tracing and 3D intersection based on the actual refracted light paths in air, a transparent observation window, and water media, ultimately completing the global transformation of multiple sets of local 3D data, the fusion of overlapping areas, and displacement calculation to obtain a continuous 360° 3D displacement field on the cylinder surface.

[0011] The technical solution adopted in this invention is as follows: A method for measuring the 360° multi-binocular collaborative displacement field of a cylinder surface in an underwater environment of a pressure vessel includes the following steps: S1. A random speckle field is prepared on the surface of the cylindrical structure to be measured, and multiple rings are set on the random speckle field. The rings and the random speckle regions retained inside them together constitute the ring-coded marker points, forming a ring-composite speckle field that can be used for DIC measurement and identification and positioning of the coded marker points. S2. Arrange the two cameras in each binocular subsystem vertically, so that the binocular baseline formed by the line connecting the optical centers of the two cameras is parallel or approximately parallel to the axis of the cylindrical structure to be measured; determine the number of binocular subsystems required to cover 360° of the cylindrical surface based on the conservative circumferential coverage angle of the cylindrical surface corresponding to a single binocular subsystem and the circumferential overlap rate of adjacent measurement areas, and arrange them in parallel along the circumference of the cylindrical structure to be measured. S3. Unify the clocks of all cameras to the same time base so that all cameras can collect data synchronously at the specified time; after the images collected by each camera are collected by the underwater data aggregation node, they are transmitted to the external equipment of the pressure cylinder through the through-tank connector. S4. Before measurement, use photogrammetry to obtain the three-dimensional coordinates and identification codes of the center of each ring-coded marker point in the global coordinate system; during measurement, perform ring detection and internal speckle matching on the upper and lower camera images acquired by each binocular subsystem, identify the identification codes of the ring-coded marker points, and obtain the three-dimensional coordinates of the center of the same marker point in the binocular local coordinate system. S5. Establish coordinate constraints using the local 3D coordinates and global 3D coordinates of the center of the same-named circular coding mark point, and obtain the initial rotation matrix and translation vector from each binocular local coordinate system to the global coordinate system; use the common cylinder under the initial pose fitting reference state to jointly optimize the coding mark point coordinate error and the radial error of the cylinder surface point to obtain the optimized rotation matrix and translation vector from each binocular local coordinate system to the global coordinate system. S6. Based on the air-transparent observation window-water medium refraction optical path model, refraction correction and ray intersection are performed on the corresponding image points in the upper and lower cameras to reconstruct the three-dimensional coordinates of the measured point in the local coordinate system of the binocular subsystem. S7. Using the optimized rotation matrix and translation vector, the local three-dimensional coordinates obtained by each binocular subsystem are transformed to the global coordinate system. The three-dimensional measurement results of the overlapping area of ​​adjacent binocular subsystems are stitched together and fused. Based on the difference between the global three-dimensional coordinates of the same surface measurement point under pressure and reference conditions, the 360° three-dimensional displacement field of the cylindrical structure surface to be measured is obtained.

[0012] Its further technical solution lies in: In S2, the method for determining the vertical arrangement and the number of binocular subsystems is as follows: the two cameras in each binocular subsystem are arranged vertically, and the optical center line of the two cameras forms the binocular baseline. The binocular baseline is parallel or approximately parallel to the axis of the cylindrical structure to be measured. The main optical axes of the two cameras are respectively oriented towards the common measurement area of ​​the binocular system, and the two cameras have a convergence angle. The vertical arrangement places the two cameras in the same group in approximately the same cylindrical circumferential observation orientation, thereby increasing the common effective measurement range of the two cameras in the cylindrical circumferential direction and improving the stability of speckle matching. Based on the horizontal coverage range of a single binocular subsystem, its corresponding circumferential coverage angle on the surface of the cylindrical structure under test is calculated. The number of binocular subsystems is then determined based on the circumferential coverage angle. The horizontal half-field of view of the camera on the air side is:

[0013] In the formula, S x The effective horizontal dimensions of the photosensitive surfaces of the upper and lower cameras; f Both refer to the effective focal length of the camera lens and use the same unit of length. c aThe horizontal half-field of view of the camera on the air side; the upper and lower cameras in the same group use the same photosensitive target size and focal length; The horizontal half-field of view after the light enters the water medium is:

[0014] In the formula, n a is the refractive index of air. n w is the refractive index of the water medium; c w The horizontal half-field of view after light enters the water medium; The midpoint of the binocular baseline of the upper and lower cameras in the same group is denoted as... O Within the cylindrical area covered by the common effective field of view of the binoculars, select the distance point. O Nearest cylinder point P , will point O To the point P The distance is used as a conservative working distance:

[0015] In the formula, O is the midpoint of the binocular baseline. O The three-dimensional coordinates; P is the nearest cylindrical point. P 3D coordinates; Z The shortest distance from the midpoint of the binocular baseline to the common effective observation cylinder of the binoculars; symbol To calculate the Euclidean norm sign of a vector; Passing point P Draw a tangent plane on the surface of the cylindrical structure to be measured, and take the direction perpendicular to the generatrix of the cylinder within this tangent plane as the circumferential tangent direction of the cylinder; the horizontal coverage width calculated according to the conservative working distance is:

[0016] In the formula, l x For the binocular subsystem at the nearest cylindrical point P Conservative horizontal coverage width along the tangential direction of the cylinder circumference; The conservative horizontal coverage width is converted into the circumferential coverage angle corresponding to the surface of the cylindrical structure to be tested:

[0017] In the formula, R The radius of the cylindrical structure to be measured; i The conservative circumferential coverage angle of a single binocular subsystem on the surface of the cylindrical structure under test; The circumferential overlap rate in the measurement areas of adjacent binocular subsystems is or At that time, the effective circumferential coverage angle of a single binocular subsystem is:

[0018] In the formula, or The circumferential overlap rate of the measurement regions of adjacent binocular subsystems, and ; i e The effective circumferential coverage angle after deducting the circumferential overlapping area; The number of binocular subsystems required to cover the entire circumference of the cylindrical structure under test is:

[0019] In the formula, N The number of binocular subsystems; The rounding symbol is used for rounding up; angles in the formula are expressed in radians. The above calculation uses the shortest distance from the midpoint of the binocular baseline to the common effective observation cylinder of the binoculars to determine the horizontal coverage width. Therefore, the number of binocular subsystems obtained is a conservative estimate to meet the 360° coverage requirement.

[0020] Multiple rings are set on a continuous random speckle field, so that a single ring and the random speckle region reserved inside it together constitute a ring-coded marker point; the ring itself is used for marker point localization, the random speckle texture inside the ring is used for identity encoding and recognition, and at the same time serves as an effective speckle for digital image correlation measurement. At a predetermined working distance and imaging resolution, 15 to 30 random speckle patterns are retained inside the ring; the imaging diameter of the random speckle patterns is 4 to 6 pixels, and the imaging linewidth of the ring is consistent with the imaging diameter of the random speckle patterns, both being 4 to 6 pixels; different ring-coded marker points have different random speckle texture distributions inside, and each random speckle texture distribution establishes a unique correspondence with the identity code of the corresponding marker point. When identifying identically coded markers in two images, the known coded marker in the reference image is used. a Using the nth circular coded marker point as a reference, compare it with the nth circular coded marker point in the image to be recognized. b Compare the candidate circular coded markers; calculate the nth candidate circular coded marker in the reference image. a The speckle sub-region inside the first circular coded marker and the first in the image to be identified b The average gray level and gray level difference normalization factor of the speckle sub-regions within the candidate circular coded markers, for the first candidate circular coded marker in the reference image. a One marker point:

[0021] In the formula, Ω represents the speckle sub-region inside the ring that participates in the matching; |Ω| represents the number of pixels contained in the speckle sub-region; and x represents the pixel coordinates of the speckle sub-region inside the ring. Fa (x) represents the gray value at this coordinate; The average gray level of the reference speckle sub-region; It is the gray-level deviation normalization factor, which is the square root of the sum of squares of the deviations of each pixel's gray level from the average gray level in the speckle sub-region. It is used to eliminate the influence of gray-level contrast on the matching results. For the first in the image to be identified b Candidate markers:

[0022] In the formula, G b Indicates the first [item] in the target image b The grayscale function of the speckle sub-region within each candidate marker point; δ is the two-dimensional sub-pixel offset of the candidate speckle sub-region relative to the initial position; The average gray level of the target speckle sub-region after offset; The gray-level deviation normalization factor for the target speckle sub-region; Define the first reference image above a The speckle sub-region inside the circular coded marker point and the first in the target image b The zero-mean normalized minimum squared distance (ZNSSD) between the speckle subregions within each candidate circular ring coded marker is:

[0023] In the formula, The ZNSSD value is used as a reference between the speckle sub-region and the candidate speckle sub-region; the smaller the ZNSSD value, the more similar the random speckle textures inside the two rings are. The sequence number of all candidate circular coded marker points in the image to be identified b Find the candidate marker point and subpixel offset that minimize the ZNSSD value from the subpixel offset δ:

[0024] In the formula, arg min represents the sequence number of all candidate coding markers. b Finding the sub-pixel offset δ C ab The combination of variables that achieves the minimum value; The index of the best matching candidate marker; The subpixel offset that minimizes the ZNSSD value; when the minimum ZNSSD value is less than a set matching threshold, the first... The circular coded marker point is identified as being related to the first... a Each of the circular coded markers has the same identity code; The center position of the circular coding marker point is determined by fitting the contours of the inner and outer edges of the circular ring to obtain the two circular ring centers c. b,1 and c b,2 The average position of the centers of the two rings is used as the initial center of the marker point; it also serves as the initial center position for extracting the speckle sub-regions inside the rings and for the reverse combined Gaussian-Newton sub-pixel iteration; the sub-pixel offset that minimizes the ZNSSD value is obtained through the reverse combined Gaussian-Newton iteration, thus obtaining the sub-pixel matching center c of the speckle sub-region. b,s The final center of the circular coded marker point is:

[0025] In the formula, c b,1 and c b,2 c is the fitting center of the inner and outer edge contours of the ring; b,s For the sub-pixel center of the speckle sub-region; c b This is the final center of the circular coding marker point.

[0026] In S4 and S5, the global coordinate unification method is as follows: The three-dimensional coordinates of the center of each circular coded marker point in the global coordinate system are obtained in advance using photogrammetry, while simultaneously determining the identity code; for the... i Images acquired by the upper and lower cameras from the binocular subsystems are used for identity encoding and recognition, and binocular reconstruction is performed to obtain the center of the corresponding marker point at the [missing information]. i The three-dimensional coordinates in the binocular local coordinate system; based on the correspondence between the local three-dimensional coordinates and the global three-dimensional coordinates of the corresponding marker points, calculate the... i The initial relative pose of the binocular local coordinate system relative to the global coordinate system is then fused with the coordinate constraints of the encoded marker points and the geometric constraints of the reference state cylinder to jointly optimize the relative pose. First, establish the coordinate error function of the marker points based on the corresponding circular coded marker points:

[0027] In the formula, i This refers to the sequence number of the binocular subsystem; K i For the first i The set of identically named circular coded marker points identified by the binocular subsystem; For the first k The global three-dimensional coordinates of the center of each marker point; For the same marker center passing through the first i Local three-dimensional coordinates reconstructed from a binocular subsystem; R i and t i These are the rotation matrix and translation vector from the local coordinate system to the global coordinate system, respectively. w ik Weights for the marker points;E M,i The global 3D coordinate registration error for the same-named marker points; Using at least three non-collinear circular coded markers in the local and global coordinate systems, the initial relative pose is obtained by minimizing the coordinate error function.

[0028] In the formula, and Let be the initial rotation matrix and translation vector obtained from the corresponding circular ring encoded markers, respectively. Together, they represent the first... i The initial relative pose of the binocular local coordinate system with respect to the global coordinate system; Using the initial relative pose, the first position in the reference state is... i The three-dimensional points on the cylindrical surface obtained by the binocular subsystem are initially transformed to the global coordinate system:

[0029] In the formula, For reference state, the first i The first binocular subsystem obtained the first j Local three-dimensional points on the surface of a column; These are global 3D points after the initial relative pose transformation; The three-dimensional points on the reference state cylinder surface after the initial conversion of each binocular subsystem are collected, and a common cylinder is fitted by finding the parameter combination that minimizes the weighted radial error from all three-dimensional points on the cylinder surface to the fitted cylinder surface.

[0030] In the formula, a is the unit direction vector of the common cylindrical axis. T Let c be its transpose; c is a point on the axis of the common cylinder. R c Let I be the radius of the common cylinder; let I be the third-order identity matrix. v ij Weights of three-dimensional points on the cylinder surface; This is the plane projection matrix of the axis method, used to calculate the radial distance from three-dimensional points on the cylinder surface to the common cylinder axis; After fixing the parameters of the common cylinder, the first... i The cylindrical geometric constraint error of the binocular subsystem is:

[0031] In the formula, E C,i For the first i The radial error of the three-dimensional points on the surface of the reference state cylinder relative to the surface of the common cylinder in the binocular subsystem; Joint optimization of the marker point coordinate error and the cylinder geometric constraint error involves finding, specifically, among all rotation matrices satisfying the rotation constraints and all translation vectors, a combination of variables that minimizes the sum of the marker point coordinate error and the weighted cylinder geometric constraint error.

[0032] In the formula, l c This is the weighting coefficient of the cylindrical geometric constraint error relative to the coordinate error of the marker point;

[0033] The first after joint optimization i Rotation matrix and translation vector from a local binocular coordinate system to a global coordinate system; Through the above joint optimization, a relative pose that simultaneously satisfies the global coordinate correspondence of the ring-coded marker points and the geometric consistency of the common cylinder of the reference state is finally obtained, which is used to... i The local three-dimensional measurement data of each binocular subsystem are uniformly converted to the global coordinate system; the common cylinder fitting and pose optimization only use the three-dimensional points on the cylinder surface under the uncompressed reference state.

[0034] In the underwater measurement system, a PTP master clock is set, and the upper and lower camera clocks of each binocular subsystem are set as slave clocks. The master clock exchanges time messages with each slave clock to determine and correct the clock deviation of each camera's local clock relative to the master clock, so that each camera uses the same time reference. The master clock determines the unified time of the q-th image acquisition according to the following formula:

[0035] In the formula, T 0 represents the start time of the measurement; q For the collection sequence number, and q =0,1,2,...;Δ T The time interval between two consecutive image acquisitions; T q For the first q The unified timing of each image acquisition; Each camera arrives at the calibrated local clock. T q Exposure acquisition is performed in real time, and each image is associated with the binocular subsystem number, camera number, acquisition sequence number, and acquisition timestamp. The underwater data aggregation node is encapsulated in a sealed pressure-resistant housing and is equipped with a multi-channel camera input interface, a channel buffer, a synchronous frame aggregation module, and a single high-speed output interface. The underwater data aggregation node assembles images from all cameras into a synchronous image group according to the acquisition sequence number and timestamp. After buffering, sorting, and data multiplexing, it forms a high-speed data stream, which is transmitted to the external processing equipment of the pressure cylinder through a through-chamber connector.

[0036] In S6, the three-dimensional reconstruction method considering refraction correction includes the following steps: S61. Define the function for calculating the direction of a single incident and outgoing ray, assuming the unit direction of the incident ray is v. in The unit direction of the emitted ray is v out Let the unit normal vector of the interface be n, the refractive index of the incident medium be n1, and the refractive index of the exiting medium be n2. Then:

[0037] In the formula, The function representing the direction of a single refraction, v in v is the unit direction of the incident ray. out Let n be the unit direction of the outgoing ray, and n be the unit normal vector of the plane of the observation window. n 1 represents the refractive index of the incident medium. n 2 represents the refractive index of the exit medium; S62. For any camera, assuming its intrinsic parameter matrix is ​​K, the coordinates of the distortion-corrected image point of the measured point in the camera image are: Its homogeneous pixel coordinates are Then, the initial ray unit direction on the air side corresponding to this image point in the camera coordinate system is:

[0038] In the formula, v a Let K be the initial ray unit direction on the air side, and K be the camera intrinsic parameter matrix. -1 This indicates the inversion of the intrinsic parameter matrix K. Let these be the homogeneous coordinates of the image point. u and v These are the horizontal and vertical pixel coordinates of the image point; S63. Calculate the water-side refracted rays in the respective camera coordinate systems; let the camera optical center be O, and the distance from the camera optical center to the inner surface of the transparent observation window be... l The thickness of the transparent observation window is t The unit normal vector of the transparent viewing window is n, and the air refractive index is [missing value]. n a The refractive index of the transparent observation window material is... n g The refractive index of water is n w ,but:

[0039] In the formula, Q1 is the intersection point of the light ray with the inner surface of the transparent observation window, Q2 is the intersection point of the light ray with the outer surface of the transparent observation window, and v g v is the unit direction of light rays inside the transparent observation window material.w L( is the unit direction of the refracted ray in the water medium) l ( ) represents the final calculated water-side refracted ray. l For ray parameters; S64. Based on steps S61, S62, and S63, calculate: Q 2L v wL These represent the starting point and direction of the water-side ray in the upper camera coordinate system, Q. 2R v wR Let the origin and direction of the water-side ray corresponding to the lower camera be defined in the lower camera coordinate system, respectively. Assume the local coordinate system coincides with the coordinate system of the upper camera. Utilizing the pose relationship between the two cameras within the subsystem, the origin and direction of the water-side refracted ray from the lower camera are transformed to the subsystem's local coordinate system. S Let R be the rotation matrix and translation vector from the camera coordinate system to S. S , t S ,but:

[0040] In the equation, the left side represents the starting point of the water-side refracted ray from the lower camera after the transformation. and direction The superscript S indicates a transformation to the local coordinate system of the binocular 3D measurement subsystem. S65, in the local coordinate system of the subsystem S Below, a local three-dimensional point is obtained by the intersection of refracted rays from the upper and lower water sides. When the two rays do not intersect strictly, the midpoint P of the line connecting the nearest points on the two rays is obtained by least squares. S And use it as a point for reconstructing three dimensions:

[0041] In the formula, and These are the optimal ray parameters in the least squares sense, P S To determine the coordinates of the measured point in the local coordinate system of the binocular 3D measurement subsystem S The three-dimensional coordinates below.

[0042] In S2, the circumferential overlap rate of adjacent measurement areas ranges from 10% to 30%.

[0043] The beneficial effects of this invention are as follows: (1) The present invention enables continuous measurement of the entire field and circumference of the column surface in a confined space.

[0044] This invention enables the non-contact acquisition of continuous and dense three-dimensional displacement information on the surface of a column, overcoming the problem that single-point or discrete-point measurements cannot fully reflect the distribution of local deformation. By arranging two cameras of the same binocular subsystem vertically along the axis of the column, the two cameras are positioned in the same circumferential observation orientation. This solves the problem of a small common effective field of view caused by the large difference in circumferential observation orientation between the two cameras and the influence of column curvature and self-occlusion in the traditional horizontal arrangement method. This significantly expands the common circumferential field of view of the binoculars and improves the stability of stereo matching of surface speckle. At the same time, this invention accurately determines the number of binocular subsystems through a systematic calculation method based on the camera target size, focal length, medium refractive index, conservative working distance, column radius, and circumferential overlap requirements. This allows multiple sets of binocular subsystems to form a continuous 360° measurement range, avoiding measurement blind spots due to insufficient number of systems or resource waste due to excessive number.

[0045] (2) The present invention can take into account both speckle measurement and marker identification without setting up independent calibration objects.

[0046] Annular markers are set on an existing random speckle field. The center of the marker point is determined by the edge of the annular marker, and the grayscale texture of the random speckles inside the annular marker is used to identify the identifier, allowing this speckle area to continue participating in DIC correlation matching. A composite speckle marker point simultaneously performs three functions: speckle measurement, marker point positioning, and identifier identification. This eliminates the need for additional dedicated coded markers printed on the surface to be measured, and also eliminates the need for dedicated 3D calibration objects at the measurement site. It is particularly suitable for practical measurement conditions where space is limited and reoperation is inconvenient after the camera, measuring fixture, and the structure to be measured are integrated. The random speckles inside the annular marker retain 15-30 speckles with an imaging diameter of 4-6 pixels, providing sufficient texture information for identifier identification without affecting the correlation matching accuracy of the DIC sub-region, thus achieving an organic unity of marking and measurement functions.

[0047] (3) The present invention reduces the number of synchronization lines, data interfaces and cross-cavity of multi-camera systems, and simplifies system wiring.

[0048] A master-slave clock synchronization mechanism is established using the PTP precision time protocol. Each camera corrects its local clock deviation by exchanging time messages and exposes independently at the same predetermined time. This eliminates the need for separate long-distance hardware trigger lines for each camera, avoiding the acquisition time deviation problems caused by differences in trigger distributors, repeaters, and cable lengths in traditional centralized triggering methods. Simultaneously, the data aggregation node inside the cylinder synchronously groups, buffers, and multiplexes multiple images according to their acquisition sequence number and timestamp, merging multiple camera images into a single high-speed data stream. This stream is then transmitted to the outside of the pressure cylinder via a single through-chamber connector, significantly reducing the number of through-chamber connectors, lowering the difficulty of sealing cylinder openings and reducing leakage risks, and simplifying the configuration of external data receiving interfaces.

[0049] (4) This invention improves the accuracy and stability of global coordinate unification of multiple binocular subsystems.

[0050] First, initial rotation and translation relationships for each binocular subsystem are established using composite speckle markers with global 3D coordinates. Then, a common cylinder is fitted using all 3D points on the reference cylinder surface, jointly optimizing the marker coordinate errors and the cylinder's radial geometric errors. Marker constraints provide a clear local-global coordinate correspondence, ensuring a deterministic basis for the relative poses between subsystems. Cylinder geometric constraints utilize a large-scale cylindrical point cloud to perform overall pose correction, effectively suppressing local registration deviations introduced by uneven marker distribution and limited marker quantity, while compensating for the impact of tooling processing errors, camera installation errors, and 3D reconstruction errors on pose calculation. The joint optimization of these two types of constraints ensures that the pose calculations of each binocular subsystem simultaneously satisfy both marker correspondence and cylinder geometric consistency, significantly reducing 3D data misalignment when stitching adjacent measurement areas. Furthermore, the common cylinder fitting only uses cylinder surface points under uncompressed reference conditions; deformed points under loaded conditions are not included in the fitting, avoiding interference from structural deformation on geometric constraints and ensuring that the optimized pose is accurately applicable to 3D reconstruction under loaded conditions.

[0051] (5) The present invention reduces the multi-medium refraction error and improves the consistency of the full-circumference displacement field.

[0052] Abandoning the traditional linear projection approximation in air environments, this method strictly follows Snell's law to sequentially trace the actual propagation direction of light in three media layers: air, a transparent observation window, and water. It calculates the intersection points of light rays with each medium interface layer by layer, obtaining the true refracted rays from the upper and lower cameras on the water side. The three-dimensional coordinates of the measured point are reconstructed through the spatial intersection of two water-side refracted rays, effectively eliminating systematic depth deviations introduced by neglecting refraction or making only simple approximations. After global coordinate transformation, the refraction reconstruction results of each binocular subsystem are weighted and fused within the overlapping region according to ZNSSD matching quality, camera observation angle, and ray intersection residuals. This further reduces single-system measurement errors, ensuring smooth and continuous transitions between adjacent measurement areas, ultimately obtaining a high-precision, continuous, and consistent 360° three-dimensional displacement field on the cylindrical surface.

[0053] In summary, this invention forms a complete, compact, and mutually supportive technical solution through the coordinated efforts of multiple stages, including binocular vertical arrangement and system quantity determination, composite speckle marking field, PTP synchronous acquisition and data aggregation, joint optimization of pose for marker points and cylinders, multi-medium refraction ray tracing 3D reconstruction, and weighted fusion displacement field calculation. The aforementioned technical features are not simply superimposed, but rather a systematic design for the specific constraints of full-circumference cylindrical measurement in underwater pressure cylinder environments. There are close causal relationships and synergistic effects among these features: the vertical arrangement determines the circumferential coverage capability and system quantity requirements of the binocular subsystem; the composite speckle marking field simultaneously solves the problems of marker point positioning, identification, and speckle measurement without occupying additional measurement space; PTP synchronous acquisition and data aggregation provide the foundation for spatiotemporal consistency across multiple systems; joint optimization of pose accurately stitches together the data from each subsystem; and refraction tracing reconstruction ensures the physical authenticity of the 3D coordinates. The aforementioned technical features work together to achieve non-contact, full-field, full-circumference, and high-precision measurement of the 360° three-dimensional displacement field of the cylinder surface in an underwater environment. This solves a long-standing technical problem in the field that has not been fully addressed by existing technologies, demonstrating significant technological advancement. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the overall flow of the measurement method of the present invention.

[0055] Figure 2 This is a schematic diagram of the arrangement of multiple binocular subsystems along the circumference of the cylindrical structure to be measured, the vertical arrangement of the upper and lower cameras, and the common effective measurement area of ​​the present invention.

[0056] Figure 3 This is a schematic diagram of the composite speckle marking field and composite speckle marking points of the present invention.

[0057] Figure 4 This is a schematic diagram of the multi-camera unified clock, synchronous acquisition, and data aggregation and transmission structure of the present invention.

[0058] Figure 5 This is a schematic diagram of the air-transparent observation window-water medium refraction light path and the intersection of the water-side refraction rays of the upper and lower cameras in this invention. Detailed Implementation

[0059] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0060] like Figure 1-Figure 5 As shown in this embodiment, a method for measuring the 360° multi-binocular cooperative displacement field of a cylinder surface in an underwater environment of a pressure cylinder includes the following steps: S1. A random speckle field is prepared on the surface of the cylindrical structure to be measured, and multiple rings are set on the random speckle field. The rings and the random speckle regions retained inside them together constitute the ring-coded marker points, forming a ring-composite speckle field that can be used for DIC measurement and identification and positioning of the coded marker points. S2. Arrange the two cameras in each binocular subsystem vertically, so that the binocular baseline formed by the line connecting the optical centers of the two cameras is parallel or approximately parallel to the axis of the cylindrical structure to be measured; determine the number of binocular subsystems required to cover 360° of the cylindrical surface based on the conservative circumferential coverage angle of the cylindrical surface corresponding to a single binocular subsystem and the circumferential overlap rate of adjacent measurement areas, and arrange them in parallel along the circumference of the cylindrical structure to be measured. S3. Unify the clocks of all cameras to the same time base so that all cameras can collect data synchronously at the specified time; after the images collected by each camera are collected by the underwater data aggregation node, they are transmitted to the external equipment of the pressure cylinder through the through-tank connector. S4. Before measurement, use photogrammetry to obtain the three-dimensional coordinates and identification codes of the center of each ring-coded marker point in the global coordinate system; during measurement, perform ring detection and internal speckle matching on the upper and lower camera images acquired by each binocular subsystem, identify the identification codes of the ring-coded marker points, and obtain the three-dimensional coordinates of the center of the same marker point in the binocular local coordinate system. S5. Establish coordinate constraints using the local 3D coordinates and global 3D coordinates of the center of the same-named circular coding mark point, and obtain the initial rotation matrix and translation vector from each binocular local coordinate system to the global coordinate system; use the common cylinder under the initial pose fitting reference state to jointly optimize the coding mark point coordinate error and the radial error of the cylinder surface point to obtain the optimized rotation matrix and translation vector from each binocular local coordinate system to the global coordinate system. S6. Based on the air-transparent observation window-water medium refraction optical path model, refraction correction and ray intersection are performed on the corresponding image points in the upper and lower cameras to reconstruct the three-dimensional coordinates of the measured point in the local coordinate system of the binocular subsystem. S7. Using the optimized rotation matrix and translation vector, the local three-dimensional coordinates obtained by each binocular subsystem are transformed to the global coordinate system. The three-dimensional measurement results of the overlapping area of ​​adjacent binocular subsystems are stitched together and fused. Based on the difference between the global three-dimensional coordinates of the same surface measurement point under pressure and reference conditions, the 360° three-dimensional displacement field of the cylindrical structure surface to be measured is obtained.

[0061] In S2, the method for determining the vertical arrangement and the number of binocular subsystems is as follows: the two cameras in each binocular subsystem are arranged vertically, and the optical center line of the two cameras forms the binocular baseline. The binocular baseline is parallel or approximately parallel to the axis of the cylindrical structure to be measured. The main optical axes of the two cameras are respectively oriented towards the common measurement area of ​​the binocular system, and the two cameras have a convergence angle. The vertical arrangement places the two cameras in the same group in approximately the same cylindrical circumferential observation orientation, thereby increasing the common effective measurement range of the two cameras in the cylindrical circumferential direction and improving the stability of speckle matching. Based on the horizontal coverage range of a single binocular subsystem, its corresponding circumferential coverage angle on the surface of the cylindrical structure under test is calculated. The number of binocular subsystems is then determined based on the circumferential coverage angle. The horizontal half-field of view of the camera on the air side is:

[0062] In the formula, S x The effective horizontal dimensions of the photosensitive surfaces of the upper and lower cameras; f Both refer to the effective focal length of the camera lens and use the same unit of length. c a The horizontal half-field of view of the camera on the air side; the upper and lower cameras in the same group use the same photosensitive target size and focal length; The horizontal half-field of view after the light enters the water medium is:

[0063] In the formula, n a is the refractive index of air. n w is the refractive index of the water medium; c w The horizontal half-field of view after light enters the water medium; The midpoint of the binocular baseline of the upper and lower cameras in the same group is denoted as... O Within the cylindrical area covered by the common effective field of view of the binoculars, select the distance point. O Nearest cylinder point P , will point O To the point P The distance is used as a conservative working distance:

[0064] In the formula, O is the midpoint of the binocular baseline. O The three-dimensional coordinates; P is the nearest cylindrical point. P 3D coordinates; Z The shortest distance from the midpoint of the binocular baseline to the common effective observation cylinder of the binoculars; symbol To calculate the Euclidean norm sign of a vector; Passing point P Draw a tangent plane on the surface of the cylindrical structure to be measured, and take the direction perpendicular to the generatrix of the cylinder within this tangent plane as the circumferential tangent direction of the cylinder; the horizontal coverage width calculated according to the conservative working distance is:

[0065] In the formula, l x For the binocular subsystem at the nearest cylindrical point P Conservative horizontal coverage width along the tangential direction of the cylinder circumference; The conservative horizontal coverage width is converted into the circumferential coverage angle corresponding to the surface of the cylindrical structure to be tested:

[0066] In the formula, R The radius of the cylindrical structure to be measured; i The conservative circumferential coverage angle of a single binocular subsystem on the surface of the cylindrical structure under test; The circumferential overlap rate in the measurement areas of adjacent binocular subsystems is or At that time, the effective circumferential coverage angle of a single binocular subsystem is:

[0067] In the formula, or The circumferential overlap rate of the measurement regions of adjacent binocular subsystems, and ; i e The effective circumferential coverage angle after deducting the circumferential overlapping area; The number of binocular subsystems required to cover the entire circumference of the cylindrical structure under test is:

[0068] In the formula, N The number of binocular subsystems; The rounding symbol is used for rounding up; angles in the formula are expressed in radians. The above calculation uses the shortest distance from the midpoint of the binocular baseline to the common effective observation cylinder of the binoculars to determine the horizontal coverage width. Therefore, the number of binocular subsystems obtained is a conservative estimate to meet the 360° coverage requirement.

[0069] Multiple rings are set on a continuous random speckle field, so that a single ring and the random speckle region reserved inside it together constitute a ring-coded marker point; the ring itself is used for marker point localization, the random speckle texture inside the ring is used for identity encoding and recognition, and at the same time serves as an effective speckle for digital image correlation measurement. At a predetermined working distance and imaging resolution, 15 to 30 random speckle patterns are retained inside the ring; the imaging diameter of the random speckle patterns is 4 to 6 pixels, and the imaging linewidth of the ring is consistent with the imaging diameter of the random speckle patterns, both being 4 to 6 pixels; different ring-coded marker points have different random speckle texture distributions inside, and each random speckle texture distribution establishes a unique correspondence with the identity code of the corresponding marker point. When identifying identically coded markers in two images, the known coded marker in the reference image is used. a Using the nth circular coded marker point as a reference, compare it with the nth circular coded marker point in the image to be recognized. b Compare the candidate circular coded markers; calculate the nth candidate circular coded marker in the reference image. a The speckle sub-region inside the first circular coded marker and the first in the image to be identified b The average gray level and gray level difference normalization factor of the speckle sub-regions within the candidate circular coded markers, for the first candidate circular coded marker in the reference image. a One marker point:

[0070] In the formula, Ω represents the speckle sub-region inside the ring that participates in the matching; |Ω| represents the number of pixels contained in the speckle sub-region; and x represents the pixel coordinates of the speckle sub-region inside the ring. F a (x) represents the gray value at this coordinate; The average gray level of the reference speckle sub-region; It is the gray-level deviation normalization factor, which is the square root of the sum of squares of the deviations of each pixel's gray level from the average gray level in the speckle sub-region. It is used to eliminate the influence of gray-level contrast on the matching results. For the first in the image to be identified b Candidate markers:

[0071] In the formula, G b Indicates the first [item] in the target image b The grayscale function of the speckle sub-region within each candidate marker point; δ is the two-dimensional sub-pixel offset of the candidate speckle sub-region relative to the initial position; The average gray level of the target speckle sub-region after offset; The gray-level deviation normalization factor for the target speckle sub-region; Define the first reference image above a The speckle sub-region inside the circular coded marker point and the first in the target image b The zero-mean normalized minimum squared distance (ZNSSD) between the speckle subregions within each candidate circular ring coded marker is:

[0072] In the formula, The ZNSSD value is used as a reference between the speckle sub-region and the candidate speckle sub-region; the smaller the ZNSSD value, the more similar the random speckle textures inside the two rings are. The sequence number of all candidate circular coded marker points in the image to be identified bFind the candidate marker point and subpixel offset that minimize the ZNSSD value from the subpixel offset δ:

[0073] In the formula, arg min represents the sequence number of all candidate coding markers. b Finding the sub-pixel offset δ C ab The combination of variables that achieves the minimum value; The index of the best matching candidate marker; The subpixel offset that minimizes the ZNSSD value; when the minimum ZNSSD value is less than a set matching threshold, the first... The circular coded marker point is identified as being related to the first... a Each of the circular coded markers has the same identity code; The center position of the circular coding marker point is determined by fitting the contours of the inner and outer edges of the circular ring to obtain the two circular ring centers c. b,1 and c b,2 The average position of the centers of the two rings is used as the initial center of the marker point; it also serves as the initial center position for extracting the speckle sub-regions inside the rings and for the reverse combined Gaussian-Newton sub-pixel iteration; the sub-pixel offset that minimizes the ZNSSD value is obtained through the reverse combined Gaussian-Newton iteration, thus obtaining the sub-pixel matching center c of the speckle sub-region. b,s The final center of the circular coded marker point is:

[0074] In the formula, c b,1 and c b,2 c is the fitting center of the inner and outer edge contours of the ring; b,s For the sub-pixel center of the speckle sub-region; c b This is the final center of the circular coding marker point.

[0075] In S4 and S5, the global coordinate unification method is as follows: The three-dimensional coordinates of the center of each circular coded marker point in the global coordinate system are obtained in advance using photogrammetry, while simultaneously determining the identity code; for the... i Images acquired by the upper and lower cameras from the binocular subsystems are used for identity encoding and recognition, and binocular reconstruction is performed to obtain the center of the corresponding marker point at the [missing information]. i The three-dimensional coordinates in the binocular local coordinate system; based on the correspondence between the local three-dimensional coordinates and the global three-dimensional coordinates of the corresponding marker points, calculate the... i The initial relative pose of the binocular local coordinate system relative to the global coordinate system is then fused with the coordinate constraints of the encoded marker points and the geometric constraints of the reference state cylinder to jointly optimize the relative pose. First, establish the coordinate error function of the marker points based on the corresponding circular coded marker points:

[0076] In the formula, i This refers to the sequence number of the binocular subsystem; K i For the first i The set of identically named circular coded marker points identified by the binocular subsystem; For the first k The global three-dimensional coordinates of the center of each marker point; For the same marker center passing through the first i Local three-dimensional coordinates reconstructed from a binocular subsystem; R i and t i These are the rotation matrix and translation vector from the local coordinate system to the global coordinate system, respectively. w ik Weights for the marker points; E M,i The global 3D coordinate registration error for the same-named marker points; Using at least three non-collinear circular coded markers in the local and global coordinate systems, the initial relative pose is obtained by minimizing the coordinate error function.

[0077] In the formula, and Let be the initial rotation matrix and translation vector obtained from the corresponding circular ring encoded markers, respectively. Together, they represent the first... i The initial relative pose of the binocular local coordinate system with respect to the global coordinate system; Using the initial relative pose, the first position in the reference state is... i The three-dimensional points on the cylindrical surface obtained by the binocular subsystem are initially transformed to the global coordinate system:

[0078] In the formula, For reference state, the first i The first binocular subsystem obtained the first j Local three-dimensional points on the surface of a column; These are global 3D points after the initial relative pose transformation; The three-dimensional points on the reference state cylinder surface after the initial conversion of each binocular subsystem are collected, and a common cylinder is fitted by finding the parameter combination that minimizes the weighted radial error from all three-dimensional points on the cylinder surface to the fitted cylinder surface.

[0079] In the formula, a is the unit direction vector of the common cylindrical axis. T Let c be its transpose; c is a point on the axis of the common cylinder.R c Let I be the radius of the common cylinder; let I be the third-order identity matrix. v ij Weights of three-dimensional points on the cylinder surface; This is the plane projection matrix of the axis method, used to calculate the radial distance from three-dimensional points on the cylinder surface to the common cylinder axis; After fixing the parameters of the common cylinder, the first... i The cylindrical geometric constraint error of the binocular subsystem is:

[0080] In the formula, E C,i For the first i The radial error of the three-dimensional points on the surface of the reference state cylinder relative to the surface of the common cylinder in the binocular subsystem; Joint optimization of the marker point coordinate error and the cylinder geometric constraint error involves finding, specifically, among all rotation matrices satisfying the rotation constraints and all translation vectors, a combination of variables that minimizes the sum of the marker point coordinate error and the weighted cylinder geometric constraint error.

[0081] In the formula, l c This is the weighting coefficient of the cylindrical geometric constraint error relative to the coordinate error of the marker point;

[0082] The first after joint optimization i Rotation matrix and translation vector from a local binocular coordinate system to a global coordinate system; Through the above joint optimization, a relative pose that simultaneously satisfies the global coordinate correspondence of the ring-coded marker points and the geometric consistency of the common cylinder of the reference state is finally obtained, which is used to... i The local three-dimensional measurement data of each binocular subsystem are uniformly converted to the global coordinate system; the common cylinder fitting and pose optimization only use the three-dimensional points on the cylinder surface under the uncompressed reference state.

[0083] In the underwater measurement system, a PTP master clock is set, and the upper and lower camera clocks of each binocular subsystem are set as slave clocks. The master clock exchanges time messages with each slave clock to determine and correct the clock deviation of each camera's local clock relative to the master clock, so that each camera uses the same time reference. The master clock determines the unified time of the q-th image acquisition according to the following formula:

[0084] In the formula, T 0 represents the start time of the measurement; q For the collection sequence number, and q=0,1,2,...;Δ T The time interval between two consecutive image acquisitions; T q For the first q The unified timing of each image acquisition; Each camera arrives at the calibrated local clock. T q Exposure acquisition is performed in real time, and each image is associated with the binocular subsystem number, camera number, acquisition sequence number, and acquisition timestamp. The underwater data aggregation node is encapsulated in a sealed pressure-resistant housing and is equipped with a multi-channel camera input interface, a channel buffer, a synchronous frame aggregation module, and a single high-speed output interface. The underwater data aggregation node assembles images from all cameras into a synchronous image group according to the acquisition sequence number and timestamp. After buffering, sorting, and data multiplexing, it forms a high-speed data stream, which is transmitted to the external processing equipment of the pressure cylinder through a through-chamber connector.

[0085] In S6, the three-dimensional reconstruction method considering refraction correction includes the following steps: S61. Define the function for calculating the direction of a single incident and outgoing ray, assuming the unit direction of the incident ray is v. in The unit direction of the emitted ray is v out Let the unit normal vector of the interface be n, the refractive index of the incident medium be n1, and the refractive index of the exiting medium be n2. Then:

[0086] In the formula, The function representing the direction of a single refraction, v in v is the unit direction of the incident ray. out Let n be the unit direction of the outgoing ray, and n be the unit normal vector of the plane of the observation window. n 1 represents the refractive index of the incident medium. n 2 represents the refractive index of the exit medium; S62. For any camera, assuming its intrinsic parameter matrix is ​​K, the coordinates of the distortion-corrected image point of the measured point in the camera image are: Its homogeneous pixel coordinates are Then, the initial ray unit direction on the air side corresponding to this image point in the camera coordinate system is:

[0087] In the formula, v a Let K be the initial ray unit direction on the air side, and K be the camera intrinsic parameter matrix. -1 This indicates the inversion of the intrinsic parameter matrix K. Let these be the homogeneous coordinates of the image point. u and v These are the horizontal and vertical pixel coordinates of the image point; S63. Calculate the water-side refracted rays in the respective camera coordinate systems; let the camera optical center be O, and the distance from the camera optical center to the inner surface of the transparent observation window be... l The thickness of the transparent observation window is t The unit normal vector of the transparent viewing window is n, and the air refractive index is [missing value]. n a The refractive index of the transparent observation window material is... n g The refractive index of water is n w ,but:

[0088] In the formula, Q1 is the intersection point of the light ray with the inner surface of the transparent observation window, Q2 is the intersection point of the light ray with the outer surface of the transparent observation window, and v g v is the unit direction of light rays inside the transparent observation window material. w L( is the unit direction of the refracted ray in the water medium) l ( ) represents the final calculated water-side refracted ray. l For ray parameters; S64. Based on steps S61, S62, and S63, calculate: Q 2L v wL These represent the starting point and direction of the water-side ray in the upper camera coordinate system, Q. 2R v wR Let the origin and direction of the water-side ray corresponding to the lower camera be defined in the lower camera coordinate system, respectively. Assume the local coordinate system coincides with the coordinate system of the upper camera. Utilizing the pose relationship between the two cameras within the subsystem, the origin and direction of the water-side refracted ray from the lower camera are transformed to the subsystem's local coordinate system. S Let R be the rotation matrix and translation vector from the camera coordinate system to S. S , t S ,but:

[0089] In the equation, the left side represents the starting point of the water-side refracted ray from the lower camera after the transformation. and direction The superscript S indicates a transformation to the local coordinate system of the binocular 3D measurement subsystem. S65, in the local coordinate system of the subsystem S Below, a local three-dimensional point is obtained by the intersection of refracted rays from the upper and lower water sides. When the two rays do not intersect strictly, the midpoint P of the line connecting the nearest points on the two rays is obtained by least squares. S And use it as a point for reconstructing three dimensions:

[0090] In the formula, and These are the optimal ray parameters in the least squares sense, P S To determine the coordinates of the measured point in the local coordinate system of the binocular 3D measurement subsystem S The three-dimensional coordinates below.

[0091] In S2, the circumferential overlap rate of adjacent measurement areas ranges from 10% to 30%.

[0092] Overall measuring fixture: This embodiment uses a surround-type measuring fixture, with the cylindrical structure to be measured installed in the middle of the measuring fixture, such as... Figure 2 As shown, the measuring fixture has multiple binocular mounting positions along the circumference of the cylindrical structure to be measured, with an upper camera and a lower camera at each position. The upper and lower cameras are respectively installed inside a sealed housing, with a transparent observation window on the side of the sealed housing facing the structure to be measured. The cameras, sealed housing, lighting components, data aggregation node, and related connecting cables are all fixed to the measuring fixture.

[0093] Before entering the underwater working environment, the camera's internal parameters are calibrated, the relative pose of the upper and lower cameras is calibrated, the parameters of the transparent observation window are measured, the circular composite speckle field is prepared, and the global coordinates of the circular coded marker points are measured. After completing the above work, the measuring fixture and the cylindrical structure to be measured located in its middle are entered into the underwater working environment as a whole, and no special calibration objects are placed or removed during the measurement process.

[0094] The number of binocular subsystems is determined: In this embodiment, the radius of the cylindrical structure to be measured is:

[0095] The upper and lower cameras of each binocular subsystem are arranged vertically along the axis of the cylindrical structure under test, with a baseline length of 300 mm. The baseline is parallel or approximately parallel to the axis of the cylinder. The principal optical axes of the two cameras are respectively oriented towards the common area under test, forming a convergence angle of approximately 30°.

[0096] The effective horizontal dimension of the camera target surface is:

[0097] The camera lens focal length is:

[0098] The distance from the midpoint of the binocular baseline to the nearest cylindrical point within the common effective field of view is:

[0099] The refractive indices of air and water are respectively taken as:

[0100] The horizontal half-field of view in the air is:

[0101] The equivalent horizontal half-field of view in water is:

[0102] The horizontal coverage width on the local tangent plane is:

[0103] The corresponding cylindrical circumferential coverage angle is:

[0104] Take the circumferential overlap rate:

[0105] The effective circumferential coverage angle of a single binocular subsystem is:

[0106] The required number of binocular subsystems is:

[0107] Therefore, in this embodiment, six sets of binocular subsystems are evenly arranged along the circumference of the cylindrical structure under test, using a total of 12 cameras. The circumferential spacing between adjacent binocular subsystems is 60°. The actual circumferential coverage angle of a single set of binocular subsystems is approximately 77.2°, and there is a continuous overlapping measurement area between adjacent binocular subsystems.

[0108] During installation, first adjust the baseline direction of each binocular subsystem to make it parallel or approximately parallel to the cylindrical axis; then adjust the observation directions of the upper and lower cameras respectively so that the effective field of view of the two cameras jointly covers the corresponding cylindrical area; finally, check the common measurement area of ​​adjacent binocular subsystems to ensure that there are no uncovered areas on the complete circumference.

[0109] Preparation of circular composite speckle field: Random speckle patterns are fabricated on the surface of the cylindrical structure under test, with the imaging diameter of the speckles in the camera image being approximately 5 pixels. Several circular rings are then placed on the random speckle field, each with an imaging linewidth of approximately 5 pixels. Approximately 24 random speckles are retained within each ring. Figure 3 As shown.

[0110] The edges of the rings are used to detect their positions, while the natural grayscale distribution of random speckles inside the rings forms distinguishable identity features and continues to participate in DIC-related calculations. The position and number of rings are determined based on the common effective field of view of each binocular subsystem, ensuring that each group of binocular subsystems can stably identify no fewer than 6 spatially distributed ring-coded markers.

[0111] Before measurement, photogrammetry is used to obtain the three-dimensional coordinates of the center of each ring-coded marker point in the global coordinate system, and the speckle sub-regions inside the rings with confirmed identities are saved as reference images. The global three-dimensional coordinate measurement directly uses the existing ring-coded marker points on the surface of the structure to be measured, without the need to place a separate dedicated three-dimensional calibration object inside the measurement fixture.

[0112] During measurement, the image processing device detects candidate ring regions in the image and calculates the ZNSSD value between the speckle sub-region inside the candidate ring and the reference image. The candidate ring with the smallest ZNSSD value that meets the set recognition threshold is identified as the corresponding ring coding marker point and assigned a corresponding identity code.

[0113] The inner and outer edges of the ring are fitted separately to obtain two center positions. The average position of the two center positions is used as the initial value for the center of the ring-coded marker point and the center of the IC-GN speckle sub-region. Then, the sub-pixel center of the speckle sub-region is obtained through IC-GN iteration. Finally, the average value of the two edge-fitted center and the sub-pixel center of the speckle sub-region is used as the center of the ring-coded marker point.

[0114] Synchronous data collection and aggregation: In this embodiment, the image acquisition frequency is set to 5 Hz, therefore the time interval between two adjacent acquisitions is:

[0115] The PTP master clock provides a unified time reference to the 12 cameras. Each camera corrects its local clock by exchanging PTP time messages and receives data when its local clock reaches the predetermined acquisition time. T q It was exposed at the time.

[0116] Compared to transmitting instantaneous trigger signals to 12 cameras separately from external devices, this method allows each camera to expose at a predetermined time according to a unified clock. This eliminates the need for independent long-distance trigger lines for each camera and reduces the number of trigger distributors and repeaters. Even if the data transmission delays differ between cameras, synchronized images from the same acquisition time can be reconstructed based on the acquisition sequence number and timestamp in the images. Figure 4 As shown.

[0117] Each image is recorded with the binocular subsystem number, upper or lower camera identifier, acquisition sequence number, and timestamp. The data aggregation node, located inside the pressure cylinder, has 12 camera input interfaces, a PTP synchronization interface, channel buffers, a synchronization frame aggregation module, and one high-speed output interface.

[0118] After receiving images from each camera, the data aggregation node first buffers the images according to the camera channel. Then, based on the acquisition sequence number and timestamp, it groups 12 images from the same acquisition time into a synchronous image group. The synchronous image group is then sorted, encapsulated, and reused to form a high-speed data stream, which is then transmitted to an external computer of the pressure cylinder through a through-chamber connector.

[0119] The external computer of the pressure cylinder recovers the upper and lower camera images of each binocular subsystem according to the synchronous image group number, providing image data from the same acquisition time for subsequent circular coded marker recognition, DIC matching, and 3D reconstruction.

[0120] Local coordinate system and global coordinate system 1: For each binocular subsystem, circular coded markers with the same identity code are identified from the upper and lower camera images, and their three-dimensional coordinates in the local coordinate system of the binocular subsystem are reconstructed based on the correspondence between the centers of the same-named markers.

[0121] For each binocular subsystem, select no fewer than 6 spatially distributed circular coded markers of the same name, and perform rigid body registration between their local 3D coordinates and the global 3D coordinates obtained by photogrammetry to obtain the initial rotation matrix and translation vector of the binocular subsystem relative to the global coordinate system.

[0122] Using the initial rotation matrix and translation vector, the 3D points on the cylinder surface obtained by each binocular subsystem in the reference state are transformed to the global coordinate system, and a common cylinder is fitted using the 3D points in the reference state of all binocular subsystems. Based on this, the coordinate errors of the corresponding circular ring-coded markers and the radial errors of the 3D points on the cylinder surface relative to the common cylinder are jointly optimized to obtain the final rotation matrix used by each binocular subsystem. Translation vector .

[0123] The circular coded markers provide a clearly identified local-global coordinate correspondence, while the common cylindrical geometric constraints utilize a large range of 3D points on the cylinder surface to perform overall pose correction for each binocular subsystem. The combination of these two types of constraints reduces pose deviations caused by marker distribution and local reconstruction errors when relying solely on a small number of markers for registration. The common cylinder is fitted using only the 3D points on the cylinder surface under the reference state. Deformation points under loading conditions do not participate in the common cylinder fitting to avoid the actual structural deformation being limited by the cylindrical geometric constraints.

[0124] Refraction 3D Reconstruction: In this embodiment, the normal distance from the camera's optical center to the inner surface of the transparent observation window is approximately 25 mm, the thickness of the transparent observation window is 10 mm, and the refractive indices of each medium are:

[0125] in, n a The refractive index of air. n g The refractive index of the transparent observation window material; n w is the refractive index of water.

[0126] Before measurement, determine the camera's intrinsic parameters, the relative pose between the upper and lower cameras, the normal vector of the transparent observation window plane, the distance from the camera's optical center to the inner surface of the observation window, and the thickness of the observation window.

[0127] For the corresponding image points of the upper and lower cameras obtained by DIC matching, the air-side unit ray corresponding to the image point is first calculated based on the camera's internal parameters; then the intersection point of the air-side ray with the inner surface of the transparent observation window is calculated, and the direction of the ray inside the observation window is obtained based on the refraction relationship between the air and the observation window material; then the intersection point of the ray with the outer surface of the observation window is calculated, and the water-side refracted ray is obtained based on the refraction relationship between the observation window material and the water medium.

[0128] Transform the water-side refracted ray corresponding to the lower camera to the local coordinate system of the binocular subsystem with the upper camera as the reference, so that the water-side refracted rays of the upper and lower cameras are expressed in the same coordinate system. Find the shortest distance between the two water-side refracted rays, and take the midpoint of the line connecting the closest points on the two rays as the three-dimensional coordinate of the measured point in the local coordinate system of the binocular subsystem.

[0129] When two water-side refracted rays intersect accurately, the length of the line connecting the nearest points is zero, and the reconstructed 3D point is the intersection of the two rays. When the two rays cannot intersect accurately due to image matching and parameter errors, the midpoint of the line connecting the nearest points is used as the reconstruction result in the least squares sense. Figure 5 As shown.

[0130] Global fusion and displacement field calculation: No. i Each binocular subsystem at the acquisition time t The first obtained j A local 3D point is denoted as The optimized rotation matrix and translation vector are used to transform it to the global coordinate system:

[0131] in, For the first i The first binocular subsystem measured the first j The three-dimensional coordinates of each measured point in the global coordinate system; superscript S i Indicates the first i Local coordinate system of each binocular subsystem; superscript GRepresents the global coordinate system.

[0132] For measured points that are simultaneously located within the overlapping measurement areas of multiple binocular subsystems, a weighted average is used for fusion:

[0133] in, In order to be able to measure the first j A set of binocular subsystems at each measured point; The fusion weights can be determined based on the ZNSSD matching quality, camera observation angle, and refracted ray intersection residuals. The resulting global 3D coordinates are the merged coordinates.

[0134] Loading status t Compared with reference state t Subtracting the global 3D coordinates of the corresponding point at 0, we obtain the 3D displacement of that point:

[0135] in, For the first j The measured points on the surface of the column at the time of data acquisition t Three-dimensional displacement relative to a reference state.

[0136] Perform the above-mentioned three-dimensional reconstruction, global coordinate transformation, overlapping area fusion and displacement calculation on the measured points within the complete circumference range to obtain a continuous 360° three-dimensional displacement field on the surface of the cylindrical structure to be measured.

[0137] This invention enables continuous full-field, full-circumference measurement of a column surface within a confined space. It obtains continuous, dense three-dimensional displacement information of the column surface non-contactly, overcoming the problem that single-point or discrete-point measurements cannot fully reflect the local deformation distribution. The number of binocular subsystems is determined based on camera imaging parameters, actual working distance, column radius, and circumferential overlap requirements, allowing multiple binocular subsystems to form a continuous 360° measurement range. The vertical arrangement with the baseline parallel to the column axis reduces the circumferential observation azimuth difference between two cameras in the same group and the influence of column self-occlusion, expands the shared circumferential field of view of both cameras, and improves the stability of stereo matching for surface speckle.

[0138] This invention combines speckle measurement and marker identification without the need for independent calibration. It places a ring on an existing random speckle field, uses the ring's edge to determine the marker center, and utilizes the grayscale texture of the random speckles within the ring for identification coding, allowing this speckle region to continue participating in DIC matching. The ring-composite speckle field simultaneously performs speckle measurement, marker positioning, and identification functions, making it suitable for measurement conditions where it is inconvenient to place dedicated calibration objects or attach temporary markers after the camera, measuring fixture, and structure under test are integrated.

[0139] This invention reduces the number of synchronization lines, data interfaces, and penetrations in multi-camera systems. It utilizes a unified clock to correct the local time of each camera, enabling each camera to expose independently at the same predetermined time. This reduces the need for trigger distributors, repeaters, and dedicated trigger cables required in centralized triggering methods, and minimizes acquisition time errors caused by line delay differences. The in-tube data aggregation node synchronously groups, buffers, and multiplexes multiple images according to their acquisition sequence number and timestamp, allowing multiple camera images to exit the chamber uniformly through a single penetration connector, simplifying in-tube wiring, penetration structure, and external data receiving interfaces.

[0140] This invention improves the accuracy and stability of global coordinate unification for multiple binocular subsystems. It first establishes the initial rotation and translation relationships of each binocular subsystem using coded markers with corresponding global 3D coordinates, and then performs joint optimization using the common cylindrical geometric features of the reference state cylinder surface. The marker constraints provide a clear local-global coordinate correspondence, while the cylindrical geometric constraints enable overall pose correction using a large range of cylinder surface points, reducing the impact of tooling processing errors, camera installation errors, local marker distribution, and 3D reconstruction errors on full-circle data stitching.

[0141] This invention reduces multi-medium refraction errors and improves the consistency of the full-circumference displacement field. It explicitly calculates the propagation direction and interface intersections of light rays in air, a transparent observation window, and water media. Three-dimensional points are reconstructed through the least-squares intersection of the refracted rays from the upper and lower cameras on the water side, avoiding the direct approximation of the refracted light path as a straight line. The refraction reconstruction results from each binocular subsystem are then fused within the overlapping region according to matching quality, observation angle, and ray intersection residuals after global coordinate transformation. This reduces three-dimensional misalignment between adjacent measurement areas, resulting in a continuous and consistent 360° three-dimensional displacement field on the cylindrical surface.

[0142] The above description is an explanation of the present invention and not a limitation thereof. The scope of the present invention is defined by the claims. Within the scope of protection of the present invention, any form of modification may be made.

Claims

1. A method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment with a pressure cylinder, characterized in that: The following steps are included: S1. A random speckle field is prepared on the surface of the cylindrical structure to be measured, and multiple rings are set on the random speckle field. The rings and the random speckle regions retained inside them together constitute the ring-coded marker points, forming a ring-composite speckle field that can be used for DIC measurement and identification and positioning of the coded marker points. S2. Arrange the two cameras in each binocular subsystem vertically, so that the binocular baseline formed by the line connecting the optical centers of the two cameras is parallel or approximately parallel to the axis of the cylindrical structure to be measured; determine the number of binocular subsystems required to cover 360° of the cylindrical surface based on the conservative circumferential coverage angle of the cylindrical surface corresponding to a single binocular subsystem and the circumferential overlap rate of adjacent measurement areas, and arrange them in parallel along the circumference of the cylindrical structure to be measured. S3. Unify the clocks of all cameras to the same time base so that all cameras can collect data synchronously at the specified time; after the images collected by each camera are collected by the underwater data aggregation node, they are transmitted to the external equipment of the pressure cylinder through the through-tank connector. S4. Before measurement, use photogrammetry to obtain the three-dimensional coordinates and identification codes of the center of each ring-coded marker point in the global coordinate system; during measurement, perform ring detection and internal speckle matching on the upper and lower camera images acquired by each binocular subsystem, identify the identification codes of the ring-coded marker points, and obtain the three-dimensional coordinates of the center of the same marker point in the binocular local coordinate system. S5. Establish coordinate constraints using the local 3D coordinates and global 3D coordinates of the center of the same-named circular coding mark point, and obtain the initial rotation matrix and translation vector from each binocular local coordinate system to the global coordinate system; use the common cylinder under the initial pose fitting reference state to jointly optimize the coding mark point coordinate error and the radial error of the cylinder surface point to obtain the optimized rotation matrix and translation vector from each binocular local coordinate system to the global coordinate system. S6. Based on the air-transparent observation window-water medium refraction optical path model, refraction correction and ray intersection are performed on the corresponding image points in the upper and lower cameras to reconstruct the three-dimensional coordinates of the measured point in the local coordinate system of the binocular subsystem. S7. Using the optimized rotation matrix and translation vector, the local three-dimensional coordinates obtained by each binocular subsystem are transformed to the global coordinate system. The three-dimensional measurement results of the overlapping area of ​​adjacent binocular subsystems are stitched together and fused. Based on the difference between the global three-dimensional coordinates of the same surface measurement point under pressure and reference conditions, the 360° three-dimensional displacement field of the cylindrical structure surface to be measured is obtained.

2. The method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment of a pressure cylinder as described in claim 1, characterized in that: In S2, the method for determining the vertical arrangement and the number of binocular subsystems is as follows: the two cameras in each binocular subsystem are arranged vertically, and the optical center line of the two cameras forms the binocular baseline. The binocular baseline is parallel or approximately parallel to the axis of the cylindrical structure to be measured. The main optical axes of the two cameras are respectively oriented towards the common measurement area of ​​the binocular system, and the two cameras have a convergence angle. The vertical arrangement places the two cameras in the same group in approximately the same cylindrical circumferential observation orientation, thereby increasing the common effective measurement range of the two cameras in the cylindrical circumferential direction and improving the stability of speckle matching. Based on the horizontal coverage range of a single binocular subsystem, its corresponding circumferential coverage angle on the surface of the cylindrical structure under test is calculated. The number of binocular subsystems is then determined based on the circumferential coverage angle. The horizontal half-field of view of the camera on the air side is: In the formula, S x The effective horizontal dimensions of the photosensitive surfaces of the upper and lower cameras; f Both refer to the effective focal length of the camera lens and use the same unit of length. γ a The horizontal half-field of view of the camera on the air side; the upper and lower cameras in the same group use the same photosensitive target size and focal length; The horizontal half-field of view after the light enters the water medium is: In the formula, n a is the refractive index of air. n w is the refractive index of the water medium; γ w The horizontal half-field of view after light enters the water medium; The midpoint of the binocular baseline of the upper and lower cameras in the same group is denoted as... O Within the cylindrical area covered by the common effective field of view of the binoculars, select the distance point. O Nearest cylinder point P , will point O To the point P The distance is used as a conservative working distance: In the formula, O is the midpoint of the binocular baseline. O The three-dimensional coordinates; P is the nearest cylindrical point. P 3D coordinates; Z The shortest distance from the midpoint of the binocular baseline to the common effective observation cylinder of the binoculars; symbol To calculate the Euclidean norm sign of a vector; Passing point P Draw a tangent plane on the surface of the cylindrical structure to be measured, and take the direction perpendicular to the generatrix of the cylinder within this tangent plane as the circumferential tangent direction of the cylinder; the horizontal coverage width calculated according to the conservative working distance is: In the formula, l x For the binocular subsystem at the nearest cylindrical point P Conservative horizontal coverage width along the tangential direction of the cylinder circumference; The conservative horizontal coverage width is converted into the circumferential coverage angle corresponding to the surface of the cylindrical structure to be tested: In the formula, R The radius of the cylindrical structure to be measured; θ The conservative circumferential coverage angle of a single binocular subsystem on the surface of the cylindrical structure under test; The circumferential overlap rate in the measurement areas of adjacent binocular subsystems is η At that time, the effective circumferential coverage angle of a single binocular subsystem is: In the formula, η The circumferential overlap rate of the measurement regions of adjacent binocular subsystems, and ; θ e The effective circumferential coverage angle after deducting the circumferential overlapping area; The number of binocular subsystems required to cover the entire circumference of the cylindrical structure under test is: In the formula, N The number of binocular subsystems; The rounding symbol is used for rounding up; angles in the formula are expressed in radians. The above calculation uses the shortest distance from the midpoint of the binocular baseline to the common effective observation cylinder of the binoculars to determine the horizontal coverage width. Therefore, the number of binocular subsystems obtained is a conservative estimate to meet the 360° coverage requirement.

3. The method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment of a pressure cylinder as described in claim 1, characterized in that: Multiple rings are set on a continuous random speckle field, so that a single ring and the random speckle region reserved inside it together constitute a ring-coded marker point; the ring itself is used for marker point localization, the random speckle texture inside the ring is used for identity encoding and recognition, and at the same time serves as an effective speckle for digital image correlation measurement. At a predetermined working distance and imaging resolution, 15 to 30 random speckle patterns are retained inside the ring; the imaging diameter of the random speckle patterns is 4 to 6 pixels, and the imaging linewidth of the ring is consistent with the imaging diameter of the random speckle patterns, both being 4 to 6 pixels; different ring-coded marker points have different random speckle texture distributions inside, and each random speckle texture distribution establishes a unique correspondence with the identity code of the corresponding marker point. When identifying identically coded markers in two images, the known coded marker in the reference image is used. a Using the nth circular coded marker point as a reference, compare it with the nth circular coded marker point in the image to be recognized. b Compare the candidate circular coded markers; calculate the nth candidate circular coded marker in the reference image. a The speckle sub-region inside the first circular coded marker and the first in the image to be identified b The average gray level and gray level difference normalization factor of the speckle sub-regions within the candidate circular coded markers, for the first candidate circular coded marker in the reference image. a One marker point: In the formula, Ω represents the speckle sub-region inside the ring that participates in the matching; |Ω| represents the number of pixels contained in the speckle sub-region; and x represents the pixel coordinates of the speckle sub-region inside the ring. F a (x) represents the gray value at this coordinate; The average gray level of the reference speckle sub-region; It is the gray-level deviation normalization factor, which is the square root of the sum of squares of the deviations of each pixel's gray level from the average gray level in the speckle sub-region. It is used to eliminate the influence of gray-level contrast on the matching results. For the first in the image to be identified b Candidate markers: In the formula, G b Indicates the first [item] in the target image b The grayscale function of the speckle sub-region within each candidate marker point; δ is the two-dimensional sub-pixel offset of the candidate speckle sub-region relative to the initial position; The average gray level of the target speckle sub-region after offset; The gray-level deviation normalization factor for the target speckle sub-region; Define the first reference image above a The speckle sub-region inside the circular coded marker point and the first in the target image b The zero-mean normalized minimum squared distance (ZNSSD) between the speckle subregions within each candidate circular ring coded marker is: In the formula, The ZNSSD value is used as a reference between the speckle sub-region and the candidate speckle sub-region; the smaller the ZNSSD value, the more similar the random speckle textures inside the two rings are. The sequence number of all candidate circular coded marker points in the image to be identified b Find the candidate marker point and subpixel offset that minimize the ZNSSD value from the subpixel offset δ: In the formula, arg min represents the sequence number of all candidate coding markers. b Finding the sub-pixel offset δ C ab The combination of variables that achieves the minimum value; The index of the best matching candidate marker; The subpixel offset that minimizes the ZNSSD value; when the minimum ZNSSD value is less than a set matching threshold, the first... The circular coded marker point is identified as being related to the first... a Each of the circular coded markers has the same identity code; The center position of the circular coding marker point is determined by fitting the contours of the inner and outer edges of the circular ring to obtain the two circular ring centers c. b,1 and c b,2 The average position of the centers of the two rings is used as the initial center of the marker point; it also serves as the initial center position for extracting the speckle sub-regions inside the rings and for the reverse combined Gaussian-Newton sub-pixel iteration; the sub-pixel offset that minimizes the ZNSSD value is obtained through the reverse combined Gaussian-Newton iteration, thus obtaining the sub-pixel matching center c of the speckle sub-region. b,s The final center of the circular coded marker point is: In the formula, c b,1 and c b,2 c is the fitting center of the inner and outer edge contours of the ring; b,s For the sub-pixel center of the speckle sub-region; c b This is the final center of the circular coding marker point.

4. The method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment of a pressure cylinder as described in claim 3, characterized in that: In S4 and S5, the global coordinate unification method is as follows: The three-dimensional coordinates of the center of each circular coded marker point in the global coordinate system are obtained in advance using photogrammetry, while simultaneously determining the identity code; for the... i Images acquired by the upper and lower cameras from the binocular subsystems are used for identity encoding and recognition, and binocular reconstruction is performed to obtain the center of the corresponding marker point at the [missing information]. i The three-dimensional coordinates in the binocular local coordinate system; based on the correspondence between the local three-dimensional coordinates and the global three-dimensional coordinates of the corresponding marker points, calculate the... i The initial relative pose of the binocular local coordinate system relative to the global coordinate system is then fused with the coordinate constraints of the encoded marker points and the geometric constraints of the reference state cylinder to jointly optimize the relative pose. First, establish the coordinate error function of the marker points based on the corresponding circular coded marker points: In the formula, i This refers to the sequence number of the binocular subsystem; K i For the first i The set of identically named circular coded marker points identified by the binocular subsystem; For the first k The global three-dimensional coordinates of the center of each marker point; For the same marker center passing through the first i Local three-dimensional coordinates reconstructed from a binocular subsystem; R i and t i These are the rotation matrix and translation vector from the local coordinate system to the global coordinate system, respectively. w ik Weights for the marker points; E M,i The global 3D coordinate registration error for the same-named marker points; Using at least three non-collinear circular coded markers in the local and global coordinate systems, the initial relative pose is obtained by minimizing the coordinate error function. In the formula, and Let be the initial rotation matrix and translation vector obtained from the corresponding circular ring encoded markers, respectively. Together, they represent the first... i The initial relative pose of the binocular local coordinate system with respect to the global coordinate system; Using the initial relative pose, the first position in the reference state is... i The three-dimensional points on the cylindrical surface obtained by the binocular subsystem are initially transformed to the global coordinate system: In the formula, For reference state, the first i The first binocular subsystem obtained the first j Local three-dimensional points on the surface of a column; These are global 3D points after the initial relative pose transformation; The three-dimensional points on the reference state cylinder surface after the initial conversion of each binocular subsystem are collected, and a common cylinder is fitted by finding the parameter combination that minimizes the weighted radial error from all three-dimensional points on the cylinder surface to the fitted cylinder surface. In the formula, a is the unit direction vector of the common cylindrical axis. T Let c be its transpose; c is a point on the axis of the common cylinder. R c Let I be the radius of the common cylinder; let I be the third-order identity matrix. v ij Weights of three-dimensional points on the cylinder surface; This is the plane projection matrix of the axis method, used to calculate the radial distance from three-dimensional points on the cylinder surface to the common cylinder axis; After fixing the parameters of the common cylinder, the first... i The cylindrical geometric constraint error of the binocular subsystem is: In the formula, E C,i For the first i The radial error of the three-dimensional points on the surface of the reference state cylinder relative to the surface of the common cylinder in the binocular subsystem; Joint optimization of the marker point coordinate error and the cylinder geometric constraint error involves finding, specifically, among all rotation matrices satisfying the rotation constraints and all translation vectors, a combination of variables that minimizes the sum of the marker point coordinate error and the weighted cylinder geometric constraint error. In the formula, λ c This is the weighting coefficient of the cylindrical geometric constraint error relative to the coordinate error of the marker point; The first after joint optimization i Rotation matrix and translation vector from a local binocular coordinate system to a global coordinate system; Through the above joint optimization, a relative pose that simultaneously satisfies the global coordinate correspondence of the ring-coded marker points and the geometric consistency of the common cylinder of the reference state is finally obtained, which is used to... i The local three-dimensional measurement data of each binocular subsystem are uniformly converted to the global coordinate system; the common cylinder fitting and pose optimization only use the three-dimensional points on the cylinder surface under the uncompressed reference state.

5. The method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment of a pressure cylinder as described in claim 1, characterized in that: In the underwater measurement system, a PTP master clock is set, and the upper and lower camera clocks of each binocular subsystem are set as slave clocks. The master clock exchanges time messages with each slave clock to determine and correct the clock deviation of each camera's local clock relative to the master clock, so that each camera uses the same time reference. The master clock determines the unified time of the q-th image acquisition according to the following formula: In the formula, T 0 represents the start time of the measurement; q For the collection sequence number, and q =0,1,2,...;Δ T The time interval between two consecutive image acquisitions; T q For the first q The unified timing of each image acquisition; Each camera arrives at the calibrated local clock. T q Exposure acquisition is performed in real time, and each image is associated with the binocular subsystem number, camera number, acquisition sequence number, and acquisition timestamp. The underwater data aggregation node is encapsulated in a sealed pressure-resistant housing and is equipped with a multi-channel camera input interface, a channel buffer, a synchronous frame aggregation module, and a single high-speed output interface. The underwater data aggregation node assembles images from all cameras into a synchronous image group according to the acquisition sequence number and timestamp. After buffering, sorting, and data multiplexing, it forms a high-speed data stream, which is transmitted to the external processing equipment of the pressure cylinder through a through-chamber connector.

6. The method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment of a pressure cylinder as described in claim 1, characterized in that: In S6, the three-dimensional reconstruction method considering refraction correction includes the following steps: S61. Define the function for calculating the direction of a single incident and outgoing ray, assuming the unit direction of the incident ray is v. in The unit direction of the emitted ray is v out Let the unit normal vector of the interface be n, the refractive index of the incident medium be n1, and the refractive index of the exiting medium be n2. Then: In the formula, The function representing the direction of a single refraction, v in v is the unit direction of the incident ray. out Let n be the unit direction of the outgoing ray, and n be the unit normal vector of the plane of the observation window. n 1 represents the refractive index of the incident medium. n 2 represents the refractive index of the exit medium; S62. For any camera, assuming its intrinsic parameter matrix is ​​K, the coordinates of the distortion-corrected image point of the measured point in the camera image are: Its homogeneous pixel coordinates are Then, the initial ray unit direction on the air side corresponding to this image point in the camera coordinate system is: In the formula, v a Let K be the initial ray unit direction on the air side, and K be the camera intrinsic parameter matrix. -1 This indicates the inversion of the intrinsic parameter matrix K. Let these be the homogeneous coordinates of the image point. u and v These are the horizontal and vertical pixel coordinates of the image point; S63. Calculate the water-side refracted rays in the respective camera coordinate systems; let the camera optical center be O, and the distance from the camera optical center to the inner surface of the transparent observation window be... l The thickness of the transparent observation window is t The unit normal vector of the transparent viewing window is n, and the air refractive index is [missing value]. n a The refractive index of the transparent observation window material is... n g The refractive index of water is n w ,but: In the formula, Q1 is the intersection point of the light ray with the inner surface of the transparent observation window, Q2 is the intersection point of the light ray with the outer surface of the transparent observation window, and v g v is the unit direction of light rays inside the transparent observation window material. w L( is the unit direction of the refracted ray in the water medium) λ ( ) represents the final calculated water-side refracted ray. λ For ray parameters; S64. Based on steps S61, S62, and S63, calculate: Q 2L v wL These represent the starting point and direction of the water-side ray in the upper camera coordinate system, Q. 2R v wR Let the origin and direction of the water-side ray corresponding to the lower camera be defined in the lower camera coordinate system, respectively. Assume the local coordinate system coincides with the coordinate system of the upper camera. Utilizing the pose relationship between the two cameras within the subsystem, the origin and direction of the water-side refracted ray from the lower camera are transformed to the subsystem's local coordinate system. S Let R be the rotation matrix and translation vector from the camera coordinate system to S. S , t S ,but: In the equation, the left side represents the starting point of the water-side refracted ray from the lower camera after the transformation. and direction The superscript S indicates a transformation to the local coordinate system of the binocular 3D measurement subsystem. S65, in the local coordinate system of the subsystem S Below, a local three-dimensional point is obtained by the intersection of refracted rays from the upper and lower water sides. When the two rays do not intersect strictly, the midpoint P of the line connecting the nearest points on the two rays is obtained by least squares. S And use it as a point for reconstructing three dimensions: In the formula, and These are the optimal ray parameters in the least squares sense, P S To determine the coordinates of the measured point in the local coordinate system of the binocular 3D measurement subsystem S The three-dimensional coordinates below.

7. The method for measuring the 360° multi-binocular collaborative displacement field of a cylindrical surface in an underwater environment of a pressure cylinder as described in claim 1, characterized in that: In S2, the circumferential overlap rate of adjacent measurement areas ranges from 10% to 30%.

Images