A non-coupling type three-dimensional detection system and method for precast beam end spraying quality

Abstract

The present application relates to the technical field of spray quality detection, and particularly relates to a precast beam end spray quality non-cooperation type three-dimensional detection system and method. The present application builds a three-dimensional digital model of the precast beam end spray surface, and matches the theoretical design surface to generate a spatial deviation field containing the coating thickness data of each surface point, so as to realize the evaluation of the thickness uniformity of the entire spray surface, and solve the problems that the prior art is difficult to comprehensively evaluate the thickness uniformity and is easy to miss local defects due to data dispersion.

Description

Technical Field

[0001] This invention relates to the field of coating quality inspection technology, specifically to a non-contact three-dimensional inspection system and method for coating quality inspection at the ends of precast beams. Background Technology

[0002] As core load-bearing components of large infrastructure such as bridges and buildings, precast beams typically require anti-corrosion coatings at their ends to resist harsh environmental erosion and ensure long-term safe service. The quality of the coating, especially the uniformity and integrity of its thickness distribution, directly affects the durability and service life of the precast beam.

[0003] In the prior art, such as CN117517325B, there is a machine vision-based aluminum single-panel spraying quality inspection and analysis system. This system acquires surface images through a camera, analyzes pits and missed spraying, and uses a thickness gauge to perform contact thickness measurement on each divided sub-region, thereby calculating parameters such as coating thickness uniformity, and finally comprehensively evaluating the spraying quality.

[0004] However, after in-depth analysis, the existing technologies still have the following shortcomings: 1. Although local thickness can be obtained, the data is discrete and cannot form a continuous thickness field covering the entire sprayed surface, making it easy to miss local defects and difficult to effectively assess the uniformity of thickness distribution; 2. There is a lack of quantitative analysis on the spatial variation and distribution continuity of thickness, making it difficult to distinguish between process defects such as uneven spraying and thickness changes caused by material rheological behavior, which in turn affects the accuracy of defect attribution judgment. At the same time, there is a lack of visual expression of the spatial location, range and type of defects, which is not conducive to guiding subsequent repair and process adjustment. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and solve the problems that existing detection methods cannot form a continuous thickness field, lack quantitative analysis of the spatial change rate and distribution continuity of coating thickness, making it difficult to effectively distinguish between process defects and thickness changes caused by coating rheological behavior, and lack a visual representation of defect type, location and range.

[0006] The technical solution adopted by the present invention to solve its technical problem is: a non-cooperative three-dimensional detection system for the coating quality of precast beam ends, including: a scanning modeling module, used to integrate multi-view data, robotic arm pose and AGV global pose to construct a three-dimensional digital model of the coating surface of the precast beam ends.

[0007] The registration deviation module is used to automatically register the 3D digital model with the theoretical design surface and generate a spatial deviation field containing coating thickness data for each surface point.

[0008] The parallel computing module is used to calculate the thickness value, local thickness gradient magnitude, and regional continuity score that characterizes the uniformity of coating thickness distribution at each surface point based on the spatial deviation field.

[0009] The logic determination module is used to perform sequential logic determination: first, identify the surface points where the local thickness gradient amplitude reaches the critical state of the coating gravity flow, and mark the points whose thickness values ​​deviate from the target value as gradient-thickness coupling defect points; then, mark the remaining surface points whose thickness values ​​deviate from the target value as pure thickness anomaly points, and mark the points whose regional continuity score is lower than that of the adjacent regions as isolated discontinuous defect points.

[0010] The defect synthesis module is used for spatial clustering and merging of all marked points to generate a spraying defect map.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention constructs a three-dimensional digital model of the sprayed surface of the precast beam end and registers it with the theoretical design surface to generate a spatial deviation field containing the coating thickness data of each surface point, thereby realizing the evaluation of the thickness uniformity of the entire sprayed surface, solving the problem that the prior art is difficult to fully evaluate the thickness uniformity due to data dispersion and is prone to missing local defects.

[0012] 2. This invention calculates the thickness value of each surface point, the local thickness gradient amplitude, and the regional continuity score that characterizes the uniformity of coating thickness distribution. Combined with the determination of the critical state of coating gravity flow, it identifies gradient-thickness coupled defect points, pure thickness anomaly points, and isolated discontinuous defect points. This solves the problem that existing technologies cannot accurately attribute the causes of defects due to the lack of quantitative analysis of spatial variation and distribution continuity.

[0013] 3. This invention generates a spraying defect map containing defect type, spatial location and area by spatial clustering and merging all marked points, providing a clear visual basis for subsequent repair and process adjustment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a schematic diagram of the system module connections of the present invention.

[0016] Figure 2 This is a schematic diagram of the detection method of the present invention.

[0017] Figure 3This is a schematic diagram illustrating the process of constructing a three-dimensional digital model for this invention.

[0018] Figure 4 This is a schematic diagram illustrating the process of identifying surface points where the local thickness gradient amplitude reaches the critical state of gravity flow of the coating, as per the present invention. Detailed Implementation

[0019] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. Furthermore, it should be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale.

[0020] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification.

[0021] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0023] The following description, in conjunction with the accompanying drawings, details the specific scheme of a non-cooperative three-dimensional inspection system for the spraying quality of precast beam ends provided by the present invention.

[0024] Please see Figure 1 The diagram shows a module connection diagram of a non-cooperative three-dimensional detection system for the coating quality of precast beam ends provided by the present invention, which specifically includes: a scanning modeling module, a registration deviation module, a parallel computing module, a logic judgment module, and a defect synthesis module.

[0025] The output of the scanning modeling module is connected to the registration deviation module, the output of the registration deviation module is connected to the parallel computing module, the input of the logic judgment module is connected to the parallel computing module, and the output is connected to the defect synthesis module.

[0026] The scanning modeling module is used to fuse multi-view data, robotic arm pose and AGV global pose to construct a three-dimensional digital model of the precast beam end sprayed surface. It mainly includes the following three steps: Step 1: Acquire multi-view data, robotic arm pose and AGV global pose.

[0027] Specifically, a 3D scanner mounted on the end of a robotic arm moves with the robotic arm to scan the sprayed surface of the precast beam from different positions and angles, thereby acquiring point cloud data from multiple perspectives, i.e., multi-view data.

[0028] By reading the joint angles output by the robotic arm controller and calculating the position and orientation of the robotic arm end effector in the robotic arm base coordinate system based on the DH parameter model, the robotic arm pose is obtained; the specific calculation process is the existing technology.

[0029] Typically, the AGV's position and orientation in the world coordinate system are output in real time through the positioning system built into the AGV, which carries the entire detection system; this is known as the AGV's global pose.

[0030] Step 2: Construct an optimization constraint network. Specifically, first, the synchronously acquired multi-view data, the robot arm pose, and the AGV global pose are timestamped and correlated, and arranged in chronological order to form the original data set.

[0031] Next, the system traverses the original data set in chronological order, using the global pose of the AGV at different times as nodes, and calculates the relative displacement and relative rotation angle between any two nodes that are consecutive in time.

[0032] The upper limit of displacement and the upper limit of rotation angle are obtained by multiplying the maximum permissible linear velocity and maximum angular velocity of the AGV by the time interval between two adjacent moments, respectively. The maximum permissible linear velocity and maximum angular velocity of the AGV are determined by the AGV equipment's technical specifications or performance parameters provided by the manufacturer.

[0033] Subsequently, based on relative displacement, upper limit of displacement, relative rotation angle and upper limit of rotation angle, the following constraints are established: (1) the relative displacement is less than or equal to the product of the upper limit of displacement and the proportional coefficient; (2) the relative rotation angle is less than or equal to the product of the upper limit of rotation angle and the proportional coefficient.

[0034] Since the time interval is known, the magnitudes of the relative displacement and relative rotation angle directly reflect the amount of pose change per unit time. Therefore, the two constraints mentioned above can together constitute a class of constraints that constrain the rate of pose change between adjacent nodes.

[0035] By combining all nodes and a set of constraints, a global motion constraint network is formed, with the overall motion trajectory of the AGV as the optimization object. This global motion constraint network constitutes the first layer of the optimization constraint network.

[0036] Then, the end-effector pose calculated using forward kinematics at each time step is used as a child node, making it subordinate to the node at the same time step. Based on the connection relationships of the robot arm's joints and real-time angle data, the coordinate transformation matrix of the robot arm's end-effector relative to its base coordinate system is calculated.

[0037] The connectivity relationships are typically described by DH parameters. (Coordinate transformation matrix) The specific calculation formula is as follows: .

[0038] in, Let be the transformation matrix of the u-th joint. The transformation matrix is ​​calculated from the DH parameters, which is a prior art method. u is the joint number, ranging from 1, 2, ..., n, where n is the total number of joints. Let be the angle of the u-th joint.

[0039] Based on this, the transformation matrix in the global coordinate system directly output by the AGV's built-in positioning system is multiplied by the coordinate transformation matrix obtained above to obtain the complex transformation matrix. The complex transformation matrix defines the initial pose relationship between the child node and its parent node.

[0040] Meanwhile, select two different time-time corresponding multi-view data, and transform them into the global coordinate system according to the initial pose of their corresponding child nodes to form point set A and point set B. Iterate the following steps through the iterative nearest point algorithm: (1) For each point in point set A, find the point with the smallest Euclidean distance in point set B as the corresponding point; (2) Based on the current correspondence, solve the rotation matrix and translation vector that minimize the sum of squared distances between corresponding points; (3) Update point set A by applying the rotation matrix and translation vector.

[0041] Repeat the above steps until the change in the average distance between corresponding points is less than a set threshold. The converged average distance between corresponding points is the spatial alignment error. Then, the spatial alignment error between multi-view data corresponding to different child nodes at different times is obtained.

[0042] After obtaining the spatial alignment error, the optimal rotation matrix and translation vector obtained by the converged iterative nearest point algorithm are combined into a rigid body transformation matrix that represents the geometric alignment relationship between the corresponding child nodes.

[0043] Using the rigid body transformation matrix as a reference, the following constraint is established: the poses of the child nodes at two corresponding moments in the global coordinate system must be equal to the relative poses calculated from the relative poses in the rigid body transformation matrix. This constraint is treated as a type II constraint on the relative poses between the corresponding child nodes.

[0044] All two types of constraints are combined with their corresponding child nodes to form a local observation constraint network. This local observation constraint network constitutes the second layer of the optimization constraint network and is sequentially associated with the first-layer global motion constraint network. Ultimately, a time-ordered, two-layer optimization constraint network is obtained.

[0045] The proportional coefficient can be set based on the AGV's repeatability, and the AGV's repeatability level is provided by the manufacturer in the product specifications. In this invention, the following correspondence applies: when the repeatability is less than 1mm, the proportional coefficient is 1-1.1; when the repeatability is 1mm-5mm, the proportional coefficient is 1.1-1.2; when the repeatability is greater than or equal to 5mm, the proportional coefficient is 1.2-1.3. The larger the repeatability value, the closer the proportional coefficient value is to the upper limit of the corresponding range.

[0046] The threshold value can typically be determined based on the ranging accuracy of the 3D scanner specified in the manufacturer's technical specifications. For example, twice the nominal ranging accuracy of the 3D scanner can be used as the threshold value; in a preferred embodiment of the present invention, the ranging accuracy of the 3D scanner used is ±0.2mm, so the threshold value can be 0.5mm.

[0047] Please see Figure 3 Step 3: Perform collaborative solution on the optimization constraint network to construct a 3D digital model. Specifically, based on the first type of constraints, the composite transformation matrix, and the second type of constraints, construct an objective function with the poses of all nodes and child nodes as optimization variables.

[0048] The specific expression for the objective function E is as follows: .

[0049] in, Let be the error term for the i-th type I constraint. If the relative displacement and relative rotation angle between adjacent nodes do not exceed the product of their upper limit and the proportional coefficient, then the error term is zero; if they do, the error term is equal to the sum of the square of the excess relative displacement and the square of the excess relative rotation angle. i is the constraint index, ranging from 1, 2, ... , This represents the total number of constraints of a certain type.

[0050] Let j be the pose error term of the j-th child node, and the pose error term is the square of the Euclidean distance between the translation part of the current pose of the child node and the translation part of the initial pose determined based on the composite transformation matrix. j is the child node index, with a value range of 1, 2, ... , This represents the total number of child nodes.

[0051] Let be the error term for the k-th type II constraint. The pose error term is obtained based on the relative pose and rigid body transformation matrix between the child nodes at corresponding two time points. Specifically, it is calculated by taking the square of the Euclidean distance between the difference of the translation vectors of the relative pose and the rigid body transformation matrix, and then adding it to the square of the angle difference between the two rotation matrices; k is the type II constraint index, with values ​​ranging from 1, 2, ... , This represents the total number of two types of constraints. The angular differences between rotation matrices can be obtained through Lie algebraic-logarithmic mapping.

[0052] α, β, and γ are all weighting coefficients greater than zero, and in this invention, they are all set to 1.5, indicating that the error terms corresponding to Class I constraints, Class II constraints, and composite transformation matrices are equally important. In practical applications, the corresponding weighting coefficients can be increased or decreased depending on the measurement accuracy of the robotic arm, AGV, and 3D scanner.

[0053] Then, based on the initial pose association defined by the composite transformation matrix, the initial pose of each child node in the global coordinate system is calculated through spatial coordinate transformation.

[0054] Using the initial pose as the initial value for iteration, a nonlinear optimization algorithm, such as the Gauss-Newton method, is employed to iteratively solve the objective function. In each iteration, the poses of all nodes and their child nodes are updated synchronously. The specific process of iteratively solving the objective function is existing technology and will not be elaborated here.

[0055] The iteration terminates when the change in the objective function value is less than the preset convergence threshold or when the maximum number of iterations is reached. At this point, the pose corresponding to each child node is the calibrated pose in the global coordinate system.

[0056] Based on this, for the multi-view data at each moment, the calibrated pose is rotated and translated to transform all the multi-view data into a global coordinate system, forming a surface observation set.

[0057] Finally, based on the surface observation set, a final 3D digital model can be generated through data fusion and 3D reconstruction using methods such as Poisson reconstruction and moving cubes. The aforementioned reconstruction methods are existing point cloud processing technologies.

[0058] The preset convergence threshold can be set to... The maximum number of iterations is set to 50. In this invention, if significant on-site vibrations lead to increased point cloud noise, the convergence threshold can be appropriately increased. The maximum number of iterations is increased to 70 to avoid getting trapped in local minima; in scenarios requiring high precision, this can be reduced to [number missing]. The maximum number of iterations remains unchanged.

[0059] The registration deviation module is used to automatically register the three-dimensional digital model with the theoretical design surface and generate a spatial deviation field characterizing the coating thickness distribution.

[0060] The specific process is as follows: First, in order to evaluate the difference between the actual sprayed surface and the theoretical design surface, the point cloud of the precast beam end sprayed surface of the three-dimensional digital model is registered with the point cloud of the theoretical design surface to obtain the initial alignment pose.

[0061] Registration can be performed using feature-based coarse registration methods, such as extracting surface curvature feature points and using RANSAC for initial alignment; the specific registration process is based on existing technologies.

[0062] Then, the iterative nearest point algorithm is executed based on the initial alignment pose to solve for the rigid body transformation matrix that aligns the sprayed surface of the precast beam end to the theoretical design surface; the specific solution process is the same as the process of calculating the spatial alignment error by the iterative nearest point algorithm.

[0063] Next, each vertex of the 3D digital model is multiplied by a rigid body transformation matrix to place it in the coordinate system of the theoretical design surface. In this coordinate system, the shortest spatial distance from each point on the sprayed surface of the precast beam end to the theoretical design surface is calculated along the surface normal direction. The shortest spatial distance is the coating thickness at that surface point.

[0064] Finally, the set of shortest spatial distances is defined as the spatial deviation field. A negative shortest spatial distance indicates that the actual surface point is within the theoretical design plane, meaning the coating is insufficient; while a positive shortest spatial distance indicates that the actual surface point is outside the theoretical design plane, meaning the coating is too thick.

[0065] The parallel computing module is used to calculate the thickness value, local thickness gradient magnitude, and regional continuity score of each surface point in parallel based on the spatial deviation field.

[0066] The specific process is as follows: the shortest spatial distance in the spatial deviation field is directly used as the thickness value of the corresponding surface point. Using the thickness value h of each surface point and its directly connected adjacent surface points, and the three-dimensional coordinates (x, y, z), the following system of linear equations is constructed: h = a × x + b × y + c × z.

[0067] Then, by solving for the coefficients a, b, and c using the least squares method, the vector (a, b, c) becomes the spatial rate of change vector of the thickness value at each surface point. The magnitude of the spatial rate of change vector is the magnitude of the local thickness gradient.

[0068] Simultaneously, a spatial spherical neighborhood with a preset radius is defined centered on each surface point, and the standard deviation of the thickness values ​​of all surface points within the spatial spherical neighborhood is calculated. The smaller the standard deviation, the more concentrated the coating thickness distribution within the spatial spherical neighborhood, and the more uniform the coating spraying at the precast beam end.

[0069] If the standard deviation is less than or equal to the zero tolerance, it can be considered that the thickness value of all surface points in the spherical neighborhood is constant, and the highest score representing complete continuity is assigned, such as 1.0.

[0070] Otherwise, the maximum standard deviation of all spatial spherical neighborhoods on the entire measured sprayed surface of the current precast beam end is used as the global normalization benchmark. The difference between the current standard deviation and the zero tolerance is divided by the difference between the global normalization benchmark and the zero tolerance, and then 1 is subtracted from the division result to obtain the regional continuity score falling within the interval [0, 1]. The closer the regional continuity score is to 0, the larger the standard deviation of the thickness value within the spatial spherical neighborhood.

[0071] In this context, a surface point and its directly connected adjacent surface points refer to two vertices directly connected by an edge in the mesh representation of a 3D digital model.

[0072] The preset radius can be determined based on the local curvature of the precast beam end. If the local curvature radius of the precast beam end is large, such as greater than 200mm, the preset radius can be 50-80mm, and in this invention, it can be 50mm. If the local curvature radius of the precast beam end is small, such as less than 50mm, the preset radius can be reduced to 30mm to avoid mixing across curved surfaces.

[0073] The zero tolerance can be determined based on the repeatability accuracy of the 3D scanner specified in the manufacturer's technical specifications. For example, if the repeatability accuracy of a certain model of 3D scanner is ±0.005 mm, then twice that, 0.01 mm, can be taken as the zero tolerance to cover the typical measurement noise range.

[0074] Furthermore, considering that at surface points with small local radii of curvature at the ends of precast beams, the surface geometry of the adjacent areas can be regarded as nonlinear surfaces, if the above linear equations are still used, the thickness changes on the nonlinear surfaces cannot be accurately characterized, resulting in errors in the calculated local thickness gradient magnitude.

[0075] Therefore, this invention introduces a curvature limit. The curvature limit is the minimum design radius of curvature selected based on the design drawings of the precast beam end.

[0076] When the local radius of curvature at the end of the precast beam is greater than or equal to the curvature limit, the spatial rate of change vector is solved using the above linear equation system; when it is less than the curvature limit, the spatial rate of change vector is solved using a quadratic surface fitting method based on the same adjacent surface points.

[0077] The local radius of curvature can be obtained by fitting the coordinates of the corresponding surface point and the adjacent surface point; the specific process of fitting the quadratic surface is an existing technology.

[0078] Please see Figure 4 The operation process of the logic determination module mainly includes the following three steps executed in sequence: Step 1: Considering that the high gradient area is prone to paint flow under the action of gravity, forming drips or exposed substrate, the surface point where the local thickness gradient amplitude reaches the critical state of paint gravity flow is identified.

[0079] The specific process is as follows: First, use the three-dimensional coordinates of all points in the spherical neighborhood corresponding to the surface point to fit the surface tangent plane using the least squares method, and then calculate the normal vector of the surface tangent plane.

[0080] Then, calculate the angle between the normal vector and the vertically downward direction, which is taken as the surface inclination angle. Multiply the gravitational acceleration by the sine of the surface inclination angle to obtain the projection component of gravitational acceleration in the tangential plane of the surface.

[0081] Next, the projected component, the density value of the coating, and the current thickness value of the corresponding surface point are multiplied to obtain the estimated shear stress on the coating at that surface point. The density value of the coating is provided by the material supplier.

[0082] If the estimated shear stress is greater than the yield stress threshold of the coating, it indicates that the shear stress on the coating at the corresponding surface point is sufficient to overcome the coating's cohesive force, and coating flow may occur. Therefore, it can be determined that the corresponding surface point has reached the critical state of gravity flow of the coating. Otherwise, it is determined that the corresponding surface point has not reached the critical state of gravity flow of the coating.

[0083] The yield stress threshold represents the minimum shear stress at which the coating begins to flow. In this invention, the nominal value of the yield stress or static shear stress specified in the technical data sheet of the coating product provided by the supplier is selected as the yield stress threshold.

[0084] Step 2: Mark the points on the surface identified in Step 1 whose thickness values ​​deviate from the target value as gradient-thickness coupling defect points.

[0085] Specifically, the thickness value deviates from the target value by offsetting the average thickness of the coating outward along its normal direction at each point on the theoretical design surface. The resulting offset surface is the target coating surface.

[0086] The Euclidean distance along the normal direction between the target coating surface and the theoretical design surface is calculated and taken as the target thickness at that point. This leads to the obtaining of the target thickness corresponding to each point on the sprayed surface of the precast beam end, with each target thickness representing the target value for that surface point.

[0087] Then, calculate the absolute value of the difference between the thickness value of each point on the sprayed surface of the precast beam end and the target value, and then divide the absolute value by the target value to obtain the relative difference.

[0088] Next, subtract the arithmetic mean of all relative differences from each relative difference to obtain the centered difference. Calculate the standard deviation of all centered differences and set the difference boundary based on this standard deviation. For example, based on statistical experience, +2σ is set as the upper limit of the difference boundary, and 2σ is set as the lower limit of the difference boundary.

[0089] If the centering difference of any surface point exceeds the difference boundary, the thickness value of that surface point is determined to deviate from the target value.

[0090] The preset average coating thickness is determined according to the engineering design requirements. For example, the thickness of the anti-corrosion coating for railway bridges is usually 2.0±0.5mm, and the specific value can be 2.0mm; if the coating is designed as a multi-layer system or has special anti-corrosion requirements, the specific value can be increased accordingly.

[0091] In a preferred embodiment of the present invention, it is also found that there is a systematic thickness deviation in the overall coating. Specifically, it is determined that the absolute value of the arithmetic mean of all relative differences is greater than the allowable systematic deviation threshold. If so, it is determined that there is a systematic thickness defect, and the arithmetic mean of all relative differences is used as the deviation amount; otherwise, it is determined that there is no systematic thickness defect.

[0092] The systematic deviation threshold can be determined based on the coating design tolerance at the end of the precast beam. For example, when the design thickness is 2.0 mm and the coating design tolerance is ±0.5 mm, the systematic deviation threshold is the ratio of the coating design tolerance to the design thickness, i.e., 0.25.

[0093] Step 3: In addition to the surface points identified in Step 1, mark the points whose thickness values ​​deviate from the target value as pure thickness anomalies among the remaining surface points.

[0094] Subsequently, based on all unmarked surface points, the average value of the regional continuity score of all surface points in the spatial spherical neighborhood corresponding to each surface point is calculated, with each surface point as the center. Points with regional continuity scores lower than the corresponding average value are marked as isolated discontinuous defect points.

[0095] The defect synthesis module is used for spatial clustering and merging of all marked points to generate a spraying defect map.

[0096] The specific process is as follows: First, all gradient-thickness coupled defect points, pure thickness anomaly points, and isolated discontinuous defect points are collected to form an initial defect point set.

[0097] Next, in the three-dimensional digital model, taking each marker point in the initial defect point set as the center, adjacent marker points with a spatial distance less than the preset merging threshold and the same defect type are merged into the same connected region.

[0098] Then, the convex hull algorithm can be used to calculate the three-dimensional boundary of the outer envelope of each connected defect region to generate independent defect elements; the specific calculation process is the existing technology.

[0099] Subsequently, the independent defect units are projected onto a two-dimensional unfolded diagram of the sprayed surface at the end of the precast beam according to their spatial location. Their location, area and defect type are marked on the unfolded diagram to form a spraying defect map.

[0100] At the same time, the judgment results of systematic thickness defects and the corresponding deviations are marked in designated areas of the two-dimensional unfolded diagram, such as the title block or legend area, so as to make the information of the spraying defect map complete.

[0101] The location can be represented by the coordinates of the polygon vertices of the independent defect element on the two-dimensional unfolded diagram, or described by the coordinate offset relative to the corner point of the precast beam. The area can be obtained by multiplying the number of pixels of the independent defect element on the two-dimensional unfolded diagram by the image resolution.

[0102] Labeling methods include, but are not limited to: using different colors to fill, text labels, and displaying area values ​​in superimposed form.

[0103] The preset merging threshold can be set according to the minimum identifiable size of the defect, typically 5-20 mm. In this invention, 10 mm is used as an example, which refers to the lower limit of resolution for manual visual inspection in industry standards. If higher detection accuracy is required or the defect size is small, the value can be appropriately reduced.

[0104] Please see Figure 2 A non-cooperative three-dimensional detection method for the coating quality of precast beam ends includes the following steps: S1, fusing multi-view data, robotic arm pose and AGV global pose to construct a three-dimensional digital model.

[0105] S2. Automatically register the 3D digital model with the theoretical design surface to generate a spatial deviation field.

[0106] S3. Based on the spatial deviation field, calculate the thickness value, local thickness gradient magnitude and regional continuity score of each surface point in parallel.

[0107] S4. Execution order logic judgment: First, mark the points on the surface that have reached the critical state of gravity flow of the coating with thickness values ​​that deviate from the target value as gradient-thickness coupling defect points; then mark the remaining points on the surface with thickness values ​​that deviate from the target value as pure thickness anomaly points, and mark the points with regional continuity scores lower than the adjacent regions as isolated discontinuous defect points.

[0108] S5. Spatial clustering and merging of all labeled points to generate a spraying defect map.

[0109] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0110] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0111] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0112] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0113] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A non-cooperative three-dimensional inspection system for precast beam end spray quality, characterized by, The application relates to a method for detecting defects of a precast beam end spraying surface, comprising the following steps: a scanning modeling module is used for fusing multi-view data, mechanical arm poses and AGV global poses to construct a three-dimensional digital model of a precast beam end spraying surface; a registration deviation module is used for automatically registering the three-dimensional digital model and a theoretical design surface to generate a spatial deviation field containing coating thickness data of each surface point; the generation process of the spatial deviation field is as follows: the precast beam end spraying surface point cloud of the three-dimensional digital model is registered with the theoretical design surface point cloud to obtain an initial alignment pose; an iterative closest point algorithm is executed with the initial alignment pose as a reference to solve a rigid transformation matrix for aligning the precast beam end spraying surface to the theoretical design surface; the three-dimensional digital model is converted to a coordinate system where the theoretical design surface is located by applying the rigid transformation matrix, and in the coordinate system, the shortest spatial distance from each point on the precast beam end spraying surface to the theoretical design surface is calculated along the surface normal direction of the point; and the set of the shortest spatial distances is defined as the spatial deviation field; a parallel computing module is used for parallel computing of thickness values, local thickness gradient amplitudes and regional continuity scores of the surface points according to the spatial deviation field; a logical judgment module is used for executing sequential logical judgment: firstly, surface points with local thickness gradient amplitudes reaching a critical state of paint gravity flow are identified, and points with thickness values deviating from target values among the surface points are marked as gradient-thickness coupling defect points; secondly, points with thickness values deviating from target values among the remaining surface points are marked as pure thickness abnormal points, and points with regional continuity scores lower than those of adjacent regions are marked as isolated discontinuous defect points; a defect comprehensive module is used for spatial clustering and fusion of all the marked points to generate a spraying defect map.

2. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 1, characterized in that, The construction process of the three-dimensional digital model is as follows: multi-view data, mechanical arm poses and AGV global poses are time-stamped, aligned and associated to form an original data set; an optimization constraint network with a double-layer structure and in time sequence is constructed based on the original data set; calibrated poses of all mechanical arm ends in a global coordinate system are obtained by collaborative calculation of the optimization constraint network; multi-view data at each time is uniformly converted to the global coordinate system based on the calibrated poses to form a surface observation set; a final three-dimensional digital model is generated by data fusion and three-dimensional reconstruction based on the surface observation set.

3. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 2, characterized in that, The optimization constraint network is as follows: the first layer is a global motion constraint network, the global poses of the AGV at different times are taken as nodes, and a class of constraints for constraining the pose change rate between adjacent nodes is established between the nodes that are continuous in time; the second layer is a local observation constraint network, the poses of the mechanical arm ends solved at each time are taken as sub-nodes, and the sub-nodes belong to the node at the same time; a composite transformation matrix is calculated according to the coordinate transformation matrix calculated based on the connection relationship and real-time angle data of each joint of the mechanical arm and the transformation matrix of the AGV in the global coordinate system; at the same time, the spatial alignment error between multi-view data at different times corresponding to different sub-nodes is calculated by an iterative closest point algorithm; and a second class of constraints for constraining the relative poses between the corresponding sub-nodes is established based on the spatial alignment error.

4. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 3, characterized in that, The process of collaborative calculation is as follows: According to a type of constraint, a composite transformation matrix and a second type of constraint, a target function is constructed with the poses of all nodes and sub-nodes as optimization variables; Based on the composite transformation matrix, the initial poses of the sub-nodes in the global coordinate system are calculated through spatial coordinate transformation; Taking the initial poses as the initial values of iteration, a nonlinear optimization algorithm is used to iteratively solve the target function, and the poses of all nodes and sub-nodes are updated synchronously in each iteration; When the change of the value of the target function is less than a preset convergence threshold or the maximum number of iterations is reached, the iteration is terminated, and the poses of the sub-nodes corresponding to each sub-node at this time are the calibrated poses in the global coordinate system.

5. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 4, characterized in that, The calculation process of the thickness value, the local thickness gradient amplitude and the regional continuity score is as follows: The shortest spatial distance in the spatial deviation field is directly taken as the thickness value of the corresponding surface point; Based on the thickness values and three-dimensional coordinates of each surface point and its adjacent surface points directly connected thereto, a spatial variation rate vector of the thickness value of each surface point is fitted and calculated, and the module length of the spatial variation rate vector is taken as the local thickness gradient amplitude; A spatial spherical neighborhood within a preset radius centered on each surface point is demarcated, and the standard deviation of the thickness values of all surface points in the spatial spherical neighborhood is calculated; If the standard deviation is less than or equal to a zero value tolerance, the highest score representing complete continuity is given; otherwise, the differences between the current standard deviation, the maximum of all standard deviations and the zero value tolerance are calculated respectively, and the regional continuity score negatively correlated with the standard deviation is calculated based on the differences.

6. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 5, characterized in that, The surface points reaching the critical state of gravity flow of the coating are identified, specifically as follows: The surface inclination angle of the surface tangent plane of the surface point is calculated according to the three-dimensional coordinates of the surface point; and the projection component of the gravitational acceleration in the surface tangent plane is determined according to the surface inclination angle; The projection component, the density value of the coating and the current thickness value of the corresponding surface point are multiplied to obtain the shear stress estimate of the coating at the surface point; If the shear stress estimate is greater than the yield stress threshold of the coating, it is determined that the corresponding surface point reaches the critical state of gravity flow of the coating.

7. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 1, characterized in that, The thickness value deviation from the target value is specifically as follows: Each point on the theoretical design surface is offset outward along its normal direction by a preset average coating thickness, and the obtained offset surface is the target coating surface; The Euclidean distance between the target coating surface and the theoretical design surface along the normal direction is calculated as the target thickness of the point; and the target thicknesses are the target values of the corresponding surface points; The absolute value of the difference between the thickness value and the target value of each point on the precast beam end spraying surface is calculated, and then the absolute value is divided by the target value to obtain the relative difference degree; The relative difference degrees are subtracted by the arithmetic mean of all relative difference degrees to obtain a group of data, and the difference degree boundary is determined based on the standard deviation of the group of data; If the relative difference degree of any surface point exceeds the difference degree boundary, it is determined that the thickness value of the surface point deviates from the target value.

8. The non-contact three-dimensional detection system for precast beam end spraying quality according to claim 1, characterized in that, The formation process of the spraying defect region is as follows: All gradient-thickness coupling defect points, pure thickness abnormal points and isolated discontinuous defect points are collected to form an initial defect point set; In the three-dimensional digital model, each marker point in the initial defect point set is taken as the center, and adjacent marker points with the same defect type and a spatial distance less than a preset merging threshold are merged into the same connected region; The three-dimensional boundary of the outer envelope of each connected defect region is calculated to generate an independent defect unit; The independent defect units are projected to a two-dimensional development diagram of the precast beam end spraying surface according to their spatial positions, and the positions, areas and defect types are marked on the development diagram to form a spraying defect atlas.

9. A non-contact 3D inspection method for precast beam end spray quality, by a non-contact 3D inspection system for precast beam end spray quality according to any one of claims 1-8, characterized by the following steps, The method comprises the following steps: S1, fusing multi-view data, mechanical arm pose and AGV global pose to construct a three-dimensional digital model; S2, automatically registering the three-dimensional digital model with the theoretical design surface to generate a spatial deviation field; S3, according to the spatial deviation field, calculating the thickness value, local thickness gradient amplitude and regional continuity score of each surface point in parallel; S4, executing sequential logic judgment: first, marking the points with thickness value deviating from the target value among the surface points reaching the critical state of paint gravity flow as gradient-thickness coupling defect points; Then, marking the points with thickness value deviating from the target value among the remaining surface points as pure thickness abnormal points, and marking the points with regional continuity score lower than that of adjacent regions as isolated discontinuous defect points; S5, spatial clustering and fusion of all marked points to generate a spraying defect atlas.

Images