A 3D vision-based method for measuring the gap width in carbon fiber composite material laying

Abstract

The application discloses a kind of based on 3D vision's carbon fiber composite material laying gap width measurement method, the gap width measurement method includes the following steps: step 1, acquisition joint area point cloud;Step 2, point cloud segmentation, point cloud is divided into target area and gap area;Step 3, gap edge point extraction, obtain target point cloud gap edge point;Step 4, four B-spline curve fitting is carried out to gap edge point;Step 5, calculate gap edge point fitting curve interval, obtain gap width minimum, maximum value;Step 6, whether the maximum value of gap width is greater than 1mm is judged.The present application utilizes actuator device to carry 3D industrial camera, can automatically position carbon fiber sheet gap, measure gap width, and further solve the problems such as high false detection rate of traditional carbon fiber sheet gap detection, low gap width measurement precision, poor reliability and the like.The method of the present application is suitable for carbon fiber composite material laying gap width measurement, and has the advantages of stability, high efficiency, precision and the like.

Description

Technical Field

[0001] This invention relates to the field of carbon fiber composite material layup gap detection, and more particularly to a method for measuring the width of carbon fiber composite material layup gaps based on 3D vision. Background Technology

[0002] The aerospace manufacturing industry is gradually transforming towards intelligent, information-based, and automated processes, with intelligent manufacturing becoming the main direction for its future. Among these processes, carbon fiber composite material layup is a crucial step in the manufacturing of aircraft components. During layup, carbon fiber composite materials cut to a specific shape are laid and assembled layer by layer on a mold according to a predetermined direction and sequence. The mold containing the multi-layered composite material is then placed in a hot press furnace for pressure, heating, and curing, ultimately forming the critical aircraft component. In this layup process, manual or mechanical layup can easily lead to inconsistent gap widths between carbon fiber sheets, thus affecting the overall performance of the component.

[0003] To address the issue of excessively wide gaps between carbon fiber sheets during the layup process, the production workshop established strict process requirements (gaps between carbon fiber sheets must be less than 1mm) and a corresponding gap width inspection plan (after each layer is laid up, workers need to visually inspect each area to ensure the gap width meets the process requirements). However, this inspection plan is not only time-consuming and labor-intensive, but also prone to causing visual fatigue for the inspectors, making it impossible to accurately and reliably measure the gap width between carbon fiber sheets. Compared with traditional manual inspection methods, 3D vision-based inspection methods offer advantages such as stability, efficiency, and accuracy. Summary of the Invention

[0004] This invention proposes a method for measuring the gap width of carbon fiber composite material layup based on 3D vision. Its main purpose is to overcome the problems of high false detection rate, low accuracy and poor reliability of traditional gap detection between material sheets.

[0005] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0006] A method for measuring the gap width of carbon fiber composite layup based on 3D vision includes the following steps:

[0007] 1) Collect point cloud data of the seam area using a 3D industrial camera, and use a region reverse search algorithm to divide the point cloud data into the target area and the gap area;

[0008] 2) The point cloud data of the target area is processed using an adaptive vector angle discrimination model to extract the gap edge points, which are then used as the gap edge points of the target point cloud;

[0009] 3) Perform a fourth-order B-spline curve fitting on the gap edge points;

[0010] 4) The curve spacing calculation model based on particle swarm optimization algorithm is used to calculate the curve spacing of the fitted curve to obtain the minimum and maximum values ​​of the gap width between the material pieces;

[0011] 5) Determine whether the maximum width of the gap between the material sheets is greater than the threshold. If it is, issue an alarm; otherwise, indicate that it is normal and continue to lay the next layer of carbon fiber composite material.

[0012] Step 1) includes the following steps:

[0013] 1.1) Construct a kd-tree for fast search of point cloud data, selecting d t As a threshold for edges in the construction of undirected graphs, multiple undirected graphs are constructed;

[0014] 1.2) Traverse the point cloud set P of the seam region to obtain the point p with minimum curvature. i As the initial seed for regional growth, the region where the seed is located is taken as the current region;

[0015] 1.3) For seed point p i k nearest neighbors N k (p i Principal component analysis is used to estimate the normal vector of the nearest neighbor point, and the angle θ between the normal vectors of the nearest neighbor point and the seed point, and the curvature difference ρ are calculated.

[0016] Point p s p j p s p j ∈N k (p i The angle θ between the normal vectors and the curvature difference ρ are defined as follows:

[0017] θ = arccos|n s T n j |

[0018] ρ=|σ s -σ j |

[0019] Where, n s n j and σ s σ j p s p j The normal vector and curvature;

[0020] 1.4) Select the threshold angle θ between the normal vectors t Curvature threshold ρ t If θ≤θ t ρ≤ρ tIf the nearest neighbor is found, add it to the current region and the seed sequence. Otherwise, perform a depth-first search on the undirected graph at the distance of the nearest neighbor. If the nearest neighbor can be found by reverse search of the current seed point, add the nearest neighbor to the current region and the seed sequence. Otherwise, do not include the nearest neighbor in the current region or add it to the seed sequence.

[0021] 1.5) Delete the current seed point and retrieve a new seed point from the seed sequence;

[0022] 1.6) Repeat steps 1.3) to 1.5) until there are no more seeds in the seed sequence. Assign labels to the data points containing the seed sequence and return to step 1.2). Continue this process until all data points have been traversed and labeled. Then, obtain the two point cloud regions with the most labels and denote them as the target point cloud regions P. seg The remaining point cloud is considered as the gap point cloud region.

[0023] Step 1.1) specifically refers to:

[0024] Let the point cloud set of the seam region be P = {p i |p i =(x i ,y i ,z i ),i∈[1,n]}, a three-dimensional spatial topological structure is established using the kd-tree model to obtain the index of the k-neighborhood, and any point p in P is selected. i , note p i The k-neighborhood is N k (p i ), p i+1 For N k (p i For any point in the array, the distance d between two points is:

[0025]

[0026] With d t As a threshold for distance to edges in an undirected graph, if d is less than or equal to d... t Then p i With p i+1 The vertices represented by p are connected by edges; otherwise, p i With p i+1 The corresponding vertices do not need to be connected by edges. Perform the above operations on the point cloud P to construct a set of distance undirected graphs G. d Represented as:

[0027]

[0028]

[0029] Among them, V i dAn undirected graph The set of vertices, An undirected graph The set of edges, m d denoted as the distance to the undirected graph, and n as the number of data points in P.

[0030] Step 1.2) specifically refers to:

[0031] Select any point p in P e Then N k (p e The covariance matrix R of ) is:

[0032]

[0033] Calculate the eigenvalue set λ of R, then the point p e curvature σ pe for:

[0034]

[0035] Calculate the curvature of all points in P, and find the point p with the minimum curvature. i for:

[0036]

[0037] Step 2) includes the following steps:

[0038] 2.1) Traverse the point cloud of the target region and its nearest neighbor p t , If the nearest neighbor set If there are gap areas, then... Add the gap edge point candidate set P can ={P i can |i∈[1,m can ]}; where, point Selecting the nearest neighbor point p within the point cloud P t m seg m is the number of data points in the target area. can The number of candidate data points for the gap edge points;

[0039] 2.2) Select set P can any point in the middle and its nearest neighbor p s , A point set is formed and fitted to a plane, then the point set is projected onto the fitted plane; where, the points... In the target area P seg Choose nearest neighbor p s ;

[0040] 2.3) with Projection point Starting from point p, the nearest neighbor point p s Projection point p s c Let s = 1…k be the endpoint, and define a vector.

[0041] 2.4) Calculate vectors Angle between the camera coordinate system and the Y-axis

[0042] 2.5) Regarding the included angle Sort by smallest to largest Then the angle α between adjacent vectors j , j = 1…k is represented as:

[0043]

[0044] 2.6) Repeat steps 2.2) to 2.5) to obtain P. can The set of angles between adjacent vectors α = {α i,j Given |i∈[1,s],j∈[1,k]}, a statistical set α is used to generate a histogram of the angle distribution between adjacent vectors, and an angle difference threshold θ is set. vec ;

[0045] 2.7) Traverse α and extract adjacent vectors whose angle is greater than θ. vec The corresponding set of data points is used as the set of gap edge points P. seam .

[0046] The edge point fitting curve expression C(u) in step 3) is:

[0047]

[0048] Where R is the set of control vertices and u is the set of node vectors.

[0049] Step 4) includes the following steps:

[0050] 4.1) In the fitted curve C(u) and T particles were randomly selected from each sample to obtain o. m and s m , with o i,m Draw a circle with center 1≤i≤T, and radius satisfying the standard for the maximum width of the gap in the manufacturing process. i,m and the part s contained in the circle j,m One particle is paired with another particle, and the pairing rule is that the shortest distance between the particles is the closest. A successfully paired particle is denoted as o. i,j,m , 1≤i≤T, 1≤j≤T;

[0051] 4.2) Repeat step 4.1) until all particles are paired and o m With s m The conditions for a one-to-one pairing are met;

[0052] 4.3) Initialize paired particles: assign initial velocities to the particles and obtain their current positions; calculate the distance f(o) between paired particles. i,j,m ), m=1, record the minimum particle spacing f(o) in the first iteration. i,j,m The corresponding paired particles are the globally optimal particle pairs G. i,j,m Other particle pairs are set as their respective individual optimal particle pairs P. i,j,m :

[0053] P i,j,m ={P i,m ,P j,m}

[0054] G i,j,m ={G i,m G j,m}

[0055] f(o i,j,m )=||o i,j,m ||

[0056] Among them, P i,j,m For paired particles o i,j,m The optimal position from initialization to the m-th iteration contains paired particles o i,j,m particles o i,m s j,m Optimal position information within m iterations; G i,j,m Let $\mathbf{m}$ be the optimal position for all paired particles from initialization up to the $m$-th iteration, which includes all paired particles $\mathbf{m}$. m s m The optimal position information within m iterations;

[0057] 4.4) In the next iteration step 4.3), update o i,j,m Based on position information and velocity, update the position G of the globally optimal particle pair. i,j,m+1 and the position P of the individual optimal particle pair i,j,m+1 :

[0058] v i,m =wv i,m-1 +b1c1(P i,m-1 -x i,m-1 )+b2c2(G i,m-1 -x i,m-1 )x i,m =x i,m-1 +a1v i,m

[0059] q j,m =wq j,m-1 +b3c3(P j,m-1 -z j,m-1 )+b4c4(G j,m-1 -z j,m-1 )z j,m =z j,m-1 +a2q j,m

[0060]

[0061]

[0062] Where b1, b2, b3, and b4 are acceleration factors; w is the inertia weight; c1, c2, c3, and c4 are uniformly distributed random numbers; and a1 and a2 are constraint factors.

[0063] 4.5) Calculate the minimum curve spacing f(G) i,j,m ), that is, the globally optimal particle pair G i,j,m The interparticle spacing f(G) i,j,m As the minimum width of the gap between the sheets, if iteration step 4.4) satisfies the condition: ||f(G) i,j,m+1 )-f(G i,j,m If || < δ or the number of iterations m ≥ K, the iteration ends; otherwise, return to step 4.4), where: δ is the error limit value, and K is the maximum number of iterations;

[0064] 4.6) Calculate the globally optimal particle pair G corresponding to the minimum curve spacing. i,j,m Direction vector and curves Paired particles s m :

[0065]

[0066] Among them, G i,m The corresponding coordinates are (G) i,m,x G i,m,y ), G j,m The corresponding coordinates are (G) j,m,x G j,m,y );

[0067] After particle o m And with The expression for parallel lines is:

[0068]

[0069] Among them, o m ={o i,m|o 1,m ,o 2,m …o T,m} represents randomly generated particles for curve C(u);

[0070] The expression for a straight line and By combining the equations, we obtain the curve. particles s m ={s j,m |s 1,m ,s 2,m …s T,m}, o m With s m Constituting paired particles o i,j,m , 1≤i≤T, 1≤j≤T;

[0071] 4.7) Repeat steps 4.3) and 4.4) to update the paired particles o i,j,m Location information and speed:

[0072] v i,m =q j,m =wv i,m-1 +b5c5(P i,m-1 -x i,m-1 )+b6c6(G i,m-1 -x i,m-1 )

[0073] z j,m =z j,m-1 +a3q j,m x i,m =x i,m-1 +a3v i,m

[0074]

[0075]

[0076] Where b5 and b6 are acceleration factors; w is the inertia weight; c5 and c6 are uniformly distributed random numbers; and a3 is the constraint factor.

[0077] 4.8) Calculate the maximum curve spacing f(G) i,j,m ), that is, the globally optimal particle pair G i,j,m The interparticle spacing f(G) i,j,m Let f(G) be the maximum width of the gap between the sheets, if it meets the condition in iteration step 4.7): ||f(G) i,j,m+1 )-f(G i,j,m If || < δ or the number of iterations m ≥ K, the iteration ends; otherwise, return to step 4.7), where: δ is the error limit value and K is the maximum number of iterations.

[0078] Step 4.1) specifically refers to:

[0079] Let the two curves fitted to the 2D edge points be represented as follows:

[0080]

[0081]

[0082] In C(u), respectively A random number of particles, T, are selected and denoted as o. m s m :

[0083] o m ={o i,m |o 1,m ,o 2,m …o T,m}

[0084] s m ={s j,m |s 1,m ,s 2,m …s T,m}

[0085] The positions of particles i and j at the m-th iteration are x and x, respectively. i,m z j,m The speeds are respectively v i,m q j,m ;

[0086] Fitted curve C(u), Particle position x m z m Curve parameter domain u m , They are represented as follows:

[0087] x m ={x i,m |x 1,m ,x 2,m …x T,m}u m ={u i,m |u 1,m ,u 2,m …u T,m}x m =C(u) m )

[0088] z m ={z j,m |z 1,m ,z 2,m …z T,m}

[0089] Fitted curve C(u), particle velocity v m q m Curve parameter offset Δu m , Represented as:

[0090] v m ={v i,m |v 1,m ,v 2,m …v T,m}Δu m ={Δu i,m |Δu 1,m ,Δu 2,m …Δu T,m}

[0091] z m ={z j,m |z 1,m ,z 2,m …z T,m}

[0092] x i,m-1 +v i,m =C(u) i,m-1 +Δu i,m )

[0093]

[0094] Successfully paired particles are denoted as o. i,j,m :

[0095] o i,j,m ={x i,m ,z j,m}

[0096] The present invention has the following beneficial effects and advantages:

[0097] 1. This invention directly uses non-contact point clouds as the basis for calculation, reducing human error generated in the traditional contact measurement process.

[0098] 2. Compared with traditional methods, the present invention has the advantages of stability, high efficiency and precision, and can effectively guarantee the performance of the produced parts. Attached Figure Description

[0099] Figure 1 This is a flowchart of a method for measuring the gap width of carbon fiber composite material layup based on 3D vision, as described in this invention.

[0100] Figure 2 This is a schematic diagram of the actuator in an embodiment of the present invention. Detailed Implementation

[0101] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0102] like Figure 1 As shown, a method for measuring the gap width of carbon fiber composite material layup based on 3D vision includes the following steps:

[0103] Step 1, Actuator Device (see...) Figure 2 The system moves the 3D industrial camera and triggers it when it reaches a fixed location to complete the point cloud acquisition of the seam area.

[0104] Step 2: Using a region reverse search algorithm, the point cloud data of the joint area is divided into the target area and the gap area.

[0105] Step 2.1: Quickly search point cloud data by constructing a kd-tree and select... As a threshold for edges in the construction of undirected graphs, to complete the construction of multiple undirected graphs.

[0106] Let the point cloud set of the seam region be P = {p i |p i =(x i ,y i ,z i For any point p in P, i ∈ [1, n], a three-dimensional spatial topology is established using a kd-tree model to implement the indexing of k-neighborhoods. i , note p i The k-neighborhood is N k (p i ), p i+1 For N k (p i For any point in the array, the distance d between two points is:

[0107]

[0108] With d t As a threshold for distance to edges in an undirected graph, if d is less than or equal to d... t Then p i With p i+1 The vertices represented by p are connected by edges; otherwise, p i With p i+1 The corresponding vertices do not need to be connected by edges. Perform the above operations on the point cloud P to construct a set of distance-based undirected graphs G. d Represented as:

[0109]

[0110]

[0111] In the formula: V i d An undirected graph The set of vertices, An undirected graph The set of edges, m d denoted as the distance to the undirected graph, and n as the number of data points in P.

[0112] Step 2.2: Traverse the point cloud set P of the seam region to obtain the point p with minimum curvature. i As the initial seed for regional growth, the region where the seed is located is taken as the current region.

[0113] Select any point p in P e Then N k (p e The covariance matrix R of ) is:

[0114]

[0115] Calculate the eigenvalue set λ of R, then the point p e curvature for:

[0116]

[0117] Calculate the curvature of all points in P, and find the point p with the minimum curvature. i for:

[0118]

[0119] Step 2.3, for seed point p i Principal component analysis is used to estimate the normal vector of the k nearest neighbor point, and the angle θ between the normal vectors of the nearest neighbor point and the seed point and the curvature difference ρ are calculated.

[0120] Let point p s p j (p s p j ∈N k (p i The angle θ between the normal vectors and the curvature difference ρ are defined as follows:

[0121] θ = arccos|n s T n j | (0.7)

[0122] ρ=|σ s -σ j | (0.8)

[0123] Where: n s n j and σ s σj p s p j The normal vector and curvature.

[0124] Step 2.4: Select the threshold angle θ between the normal vectors. t Curvature threshold ρ t If θ≤θ t ρ≤ρ t If the nearest neighbor is added to the current region, it will be added to the seed sequence.

[0125] Step 2.5: If the nearest neighbor does not meet the conditions of Step 2.4, perform a depth-first search on the undirected graph where the nearest neighbor is located. If the nearest neighbor can be found by reverse search to the current seed point, add the nearest neighbor to the current region and add it to the seed sequence; otherwise, the nearest neighbor is not included in the current region and is not added to the seed sequence.

[0126] Step 2.6: Delete the current seed point, retrieve the new seed point from the seed sequence, and return to step 2.3.

[0127] Step 2.7: Repeat this process until there are no more seeds in the seed sequence. Assign labels to the data points from the above process and return to Step 2. After traversing all data points and assigning labels, obtain the two point cloud regions with the most labels and denote them as the target point cloud regions P. seg .

[0128] Step 3: Use an adaptive vector angle discrimination model to extract the gap edge points from the target point cloud data to obtain the gap edge points of the target point cloud.

[0129] Step 3.1: Traverse the point cloud of the target region and its nearest neighbors If the nearest neighbor set If there are gap areas, then... Included in the candidate set of gap edge points Among them, point It is in point cloud P Choose nearest neighbor p t m seg m is the number of data points in the target area. can This represents the number of candidate data points for the gap edge point.

[0130] Step 3.2: Select set P can any point in the middle and its nearest neighbors A point set is formed and fitted to a plane, then the point set is projected onto that plane; where, the points... In the target area P seg Choose nearest neighbor p s .

[0131] Step 3.3, with Projection point Starting from point p, the nearest neighbor point p s Projection point Define a vector as the endpoint.

[0132] Step 3.4: Calculate the vector Angle between the camera coordinate system and the Y-axis

[0133] Step 3.5, Adjust the included angle Sort by smallest to largest Then the angle α between adjacent vectors j (j=1…k) is represented as:

[0134]

[0135] Step 3.6: Repeat steps 2-5 above to obtain P. can The set of angles between adjacent vectors α = {α i,j Given |i∈[1,s],j∈[1,k]}, a statistical set α is used to generate a histogram of the angle distribution between adjacent vectors, and the angle difference threshold θ is calculated. vec .

[0136] Step 3.7: Traverse α and extract adjacent vectors whose included angle is greater than θ. vec The corresponding set of data points is then the set of gap edge points P. seam .

[0137] Step 4: Perform a fourth-order B-spline curve fitting on the edge points of the gap.

[0138] From the formula

[0139]

[0140] It can be seen that when the curve order k = 4 and the number of control vertices is 5, the mathematical expression of the B-spline curve is:

[0141]

[0142]

[0143] Where 0≤u≤1, i=0,1,…,4,

[0144] From the formula, the mathematical expression for the edge point fitting curve can be obtained as follows:

[0145]

[0146] Where R is the set of control vertices and u is the set of node vectors.

[0147] Step 5: Use a particle swarm optimization algorithm-based curve spacing calculation model to calculate the curve spacing of the fitted curves to obtain the minimum and maximum gap widths between the material pieces.

[0148] Step 5.1, in the fitted curve C(u) and T particles were randomly selected to obtain o m and s m , with o i,m Draw a circle with center (1≤i≤T) and radius satisfying the maximum width standard for manufacturing gaps. For o... i,m and the part s contained in the circle j,m One particle is paired with another particle, and the pairing rule is that the shortest distance between the particles is the closest. A successfully paired particle is denoted as o. i,j,m (1≤i≤T, 1≤j≤T).

[0149] Let the two curves fitted to the 2D edge points be represented as follows:

[0150]

[0151] In C(u), respectively A random number of particles, T, are selected and denoted as o. m s m :

[0152]

[0153] The positions of particles i and j at the m-th iteration are x i,m z j,m speed v i,m q j,m They are represented as follows:

[0154]

[0155] Fitted curve C(u), Particle position x m z m Curve parameter domain u m , They are represented as follows:

[0156]

[0157] Fitted curve C(u), particle velocity v m q m Curve parameter offset Δu m , Represented as:

[0158]

[0159] Successfully paired particles are denoted as o. i,j,m :

[0160] o i,j,m ={x i,m ,z j,m} (0.19)

[0161] Step 5.2: Repeat step 5.1 until all particles are paired, and o m With s m The conditions for a one-to-one pairing are met.

[0162] Step 5.3: Initialize paired particles, assign initial velocities to the particles and obtain their current positions. Calculate the distance f(o) between paired particles. i,j,m (m=1), record the minimum f(o) in the first iteration. i,j,m The corresponding paired particles are the globally optimal particle pair G. i,j,m Other particle pairs are set as their respective individual optimal particle pairs P. i,j,m .

[0163] P i,j,m ={P i,m ,P j,m}

[0164] G i,j,m ={G i,m G j,m} (0.20)

[0165] f(o i,j,m )=||o i,j,m ||

[0166] In the formula: P i,j,m For paired particles o i,j,m The optimal position from initialization to the m-th iteration contains paired particles o i,j,m particles o i,m s j,m Optimal position information within m iterations; G i,j,m Let $\mathbf{m}$ be the optimal position for all paired particles from initialization up to the $m$-th iteration, which includes all particle $\mathbf{m}$. m s m The optimal position information within m iterations.

[0167] Step 5.4, Second iteration, update o i,j,m Based on position information and velocity, update the position G of the globally optimal particle pair. i,j,m+1 The optimal position P of an individual particle pair i,j,m+1 .

[0168]

[0169]

[0170]

[0171] In the formula: b1, b2, b3, b4 are acceleration factors used to adjust the convergence speed; w is the inertia weight used to adjust the global search and local search; c1, c2, c3, c4 are uniformly distributed random numbers; a1 and a2 are constraint factors, which are used to control the weights of the speed.

[0172] Step 5.5: Calculate the minimum curve spacing f(G) i,j,m ), that is, the globally optimal particle pair G i,j,m The particle spacing in the figure is used as the minimum width of the gap between the material pieces. If iteration step 5.4 satisfies the condition: ||f(G i,j,m+1 )-f(G i,j,m If || < δ or the number of iterations m ≥ K, the iteration ends; otherwise, return to step 5.4. Where: δ is the error limit value, and K is the maximum number of iterations.

[0173] Step 5.6: Calculate the paired particle G corresponding to the minimum curve spacing. i,j,m Direction vector Calculation curve Paired particles s m :

[0174] Calculate the direction vector

[0175]

[0176] In the formula: G i,j,m For the paired particles corresponding to the minimum curve spacing, G i,m The corresponding coordinates are (G) i,m,x G i,m,y ), G j,m The corresponding coordinates are (G) j,m,x G j,m,y ).

[0177] After particle o m And with The expression for parallel lines is:

[0178]

[0179] In the formula: o m ={o i,m |o 1,m ,o 2,m …o T,m} represents the random generation of particles for the curve C(u).

[0180] The expression for a straight line and Combined equations, output curve particles s m ={s j,m |s 1,m ,s 2,m …s T,m}, o m With s m Formation of paired particles o i,j,m (1≤i≤T, 1≤j≤T).

[0181] Step 5.7: Perform the initialization from Step 5.3 and the iterative update from Step 5.4, pairing particles o i,j,m The location information and speed update rules are as follows:

[0182]

[0183]

[0184]

[0185] In the formula: b5 and b6 are acceleration factors used to adjust the convergence speed; w is the inertia weight used to adjust the global search and local search; c5 and c6 are uniformly distributed random numbers; a3 is a constraint factor, which is used to control the weight of the speed.

[0186] Step 5.8: Calculate the maximum curve spacing f(G) i,j,m ), that is, the globally optimal particle pair G i,j,m The interparticle spacing f(G) i,j,m The maximum width of the gap between the sheets is defined as ||f(G) as the maximum width of the gap between the sheets, if the condition in iteration step 5.7 is satisfied: i,j,m+1 )-f(G i,j,m If || < δ or the number of iterations m ≥ K, the iteration ends; otherwise, return to step 5.7. Where: δ is the error limit value, and K is the maximum number of iterations.

[0187] Step 6: Determine whether the maximum width of the gap between the material pieces is greater than 1mm.

[0188] If f(G) i,j,m If the thickness exceeds 1mm, the system will issue an alarm; otherwise, it will indicate that everything is normal and the workshop workers will continue to lay the next layer of carbon fiber composite material.

[0189] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A method for measuring the gap width of carbon fiber composite material layup based on 3D vision, characterized in that, Includes the following steps: 1) Collect point cloud data of the seam area using a 3D industrial camera, and use a region reverse search algorithm to divide the point cloud data into the target area and the gap area; 2) The point cloud data of the target area is processed using an adaptive vector angle discrimination model to extract the gap edge points, which are then used as the gap edge points of the target point cloud; 3) Perform fourth-order B-spline curve fitting on the gap edge points; 4) The curve spacing calculation model based on particle swarm optimization algorithm is used to calculate the curve spacing of the fitted curve to obtain the minimum and maximum values ​​of the gap width between the material pieces; 5) Determine if the maximum width of the gap between the material pieces is greater than the threshold. If it is, issue an alarm. Otherwise, the message indicates normal operation, and the next layer of carbon fiber composite material is laid down. Step 4) includes the following steps: 4.1) In the fitted curve and All are randomly selected Each particle received and ,by , Draw a circle with the center as the center, and the radius meets the standard for the maximum width of the gap in the manufacturing process. and the part contained in the circle One particle is paired with another particle, and the pairing rule is that the shortest distance between the particles is the greatest. Successfully paired particles are denoted as _____. , ; 4.2) Repeat step 4.1) until all particles are paired, and and The conditions for a one-to-one pairing are met; 4.3) Initialize paired particles: assign initial velocities to the particles and obtain their current positions; calculate the distance between paired particles. , Record the minimum particle spacing in the first iteration. The corresponding paired particles are the globally optimal particle pairs. Other particle pairs are set as their respective individual optimal particle pairs. : in, For paired particles From initialization to the... The optimal position up to the next iteration, which contains paired particles. particles , exist Optimal position information within the next iteration; For all paired particles from initialization to the 1st The optimal position up to the next iteration, which contains all paired particles. , exist Optimal position information within the next iteration; 4.4) Second iteration step 4.3), update Based on position information and velocity, update the position of the globally optimal particle pair. and the position of the individual optimal particle pair : in, As an acceleration factor; Inertial weights; The numbers are uniformly distributed random numbers; Constraint factors; 4.5) Calculate the minimum curve spacing That is, the globally optimal particle pair interparticle spacing As the minimum width of the gap between the material sheets, if iteration step 4.4) satisfies the following condition: or number of iterations If the iteration ends, the iteration ends; otherwise, return to step 4.4, where: This is the error limit value. This represents the maximum number of iterations. 4.6) Calculate the globally optimal particle pair corresponding to the minimum curve spacing. Direction vector and curves Paired particles : in, Corresponding coordinates are , Corresponding coordinates are ; After particles And with The expression for parallel lines is: in, For curves Randomly generate particles; The expression for a straight line and By combining the equations, we obtain the curve. particles , and Constituting paired particles , ; 4.7) Repeat steps 4.3) and 4.4) to update the paired particles. Location information and speed: in, As an acceleration factor; Inertial weights; The numbers are uniformly distributed random numbers; Constraint factors; 4.8) Calculate the maximum curve spacing. That is, the globally optimal particle pair interparticle spacing As the maximum width of the gap between the material sheets, if the condition in iteration step 4.7 is satisfied: or number of iterations If the iteration ends, the iteration ends; otherwise, return to step 4.7, where: This is the error limit value. This represents the maximum number of iterations.

2. The method for measuring the gap width of carbon fiber composite material layup based on 3D vision according to claim 1, characterized in that, Step 1) includes the following steps: 1.1) Construct a kd-tree for fast search of point cloud data, selecting... As a threshold for edges in the construction of undirected graphs, multiple undirected graphs are constructed; 1.2) Traverse the point cloud set of the seam region To obtain the point of minimum curvature As the initial seed for regional growth, the region where the seed is located is taken as the current region; 1.3) For seed points of Nearest neighbor Principal component analysis is used to estimate the normal vectors of nearest neighbors, and the angle between the normal vectors of the nearest neighbors and the seed point is calculated. Curvature difference : point , , Angle between normal vectors Curvature difference Defined as: in, , and , They are respectively , The normal vector and curvature; 1.4) Select the threshold for the angle between the normal vectors Curvature threshold ;like , If the nearest neighbor is found, add it to the current region and the seed sequence. Otherwise, perform a depth-first search on the undirected graph at the distance of the nearest neighbor. If the nearest neighbor can be found by reverse search of the current seed point, add the nearest neighbor to the current region and the seed sequence. Otherwise, do not include the nearest neighbor in the current region or add it to the seed sequence. 1.5) Delete the current seed point and retrieve a new seed point from the seed sequence; 1.6) Repeat steps 1.3) to 1.5) until there are no more seeds in the seed sequence. Assign labels to the data points containing the seed sequence and return to step 1.2). After traversing all data points and assigning labels, obtain the two point cloud regions with the most labels and denote them as the target point cloud regions. The remaining point cloud is considered as the gap point cloud region.

3. The method for measuring the gap width of carbon fiber composite material layup based on 3D vision according to claim 2, characterized in that, Step 1.1) specifically refers to: Set the point cloud set for the seam region A three-dimensional spatial topology is established using a kd-tree model to obtain... The index of the neighborhood is selected. any point in the middle ,remember of Neighborhood is , for Any point in the array, the distance between two points for: by As a threshold for distance to edges in an undirected graph, if Less than or equal to ,but and The vertices represented are connected by edges, otherwise and The corresponding vertices do not need to be connected by edges, for point clouds Perform the above operations to construct a set of distance-based undirected graphs. Represented as: in, An undirected graph The set of vertices, An undirected graph The set of edges, The distance represents the number of undirected graphs. for Number of data points in the middle.

4. The method for measuring the gap width of carbon fiber composite material layup based on 3D vision according to claim 2, characterized in that, Step 1.2) specifically refers to: Select any point in the middle ,but covariance matrix for: calculate eigenvalue set Then point curvature for: calculate Given the curvature of all points, find the point with the minimum curvature. for: 。 5. The method for measuring the gap width of carbon fiber composite material layup based on 3D vision according to claim 1, characterized in that, Step 2) includes the following steps: 2.1) Traverse the point cloud of the target region and its nearest neighbors , If the nearest neighbor set If there are gap areas, then... Add gap edge point candidate set Among them, point It is in point cloud Select nearest neighbor , The number of data points in the target area. The number of candidate data points for the gap edge point; 2.2) Selecting a set any point in the middle and its nearest neighbors , A point set is formed and fitted to a plane, then the point set is projected onto the fitted plane; where, the points... In the target area Select nearest neighbor ; 2.3) with Projection point Starting from, nearest neighbor points Projection point , Define a vector as the endpoint. ; 2.4) Calculate vectors With camera coordinate system Angle between axes , ; 2.5) Regarding the included angle Sort by smallest to largest Then the angle between adjacent vectors , Represented as: ； 2.6) Repeat steps 2.2) to 2.5) to obtain... The set of angles between adjacent vectors Statistical set Generate a histogram of the angle distribution between adjacent vectors and set an angle difference threshold. ; 2.7) Traversal Extract the vectors whose angle is greater than 1. The corresponding set of data points is used as the set of gap edge points. .

6. The method for measuring the gap width of carbon fiber composite material layup based on 3D vision according to claim 1, characterized in that, The edge point fitting curve expression in step 3) for: in, To control the vertex set, It is a set of node vectors.

7. The method for measuring the gap width of carbon fiber composite material layup based on 3D vision according to claim 1, characterized in that, Step 4.1) specifically refers to: Let the two curves fitted to the 2D edge points be represented as follows: In respectively , Randomly select quantity The particle is denoted as , : No. , Particles in the first The positions at the next iteration are respectively , The speeds are respectively , ; Fitted curve , Particle position , Curve parameter domain , They are represented as follows: Fitted curve , particle velocity , Curve parameter offset , Represented as: Successfully paired particles are denoted as : 。

Images