Three-dimensional reconstruction method for ship virtual teaching modeling

Abstract

The invention belongs to the technical field of virtual teaching modeling methods, and particularly relates to a three-dimensional reconstruction method for ship virtual teaching modeling. Comprising the following steps: establishing a point cloud data coordinate system, optimizing ship space contour feature points, optimizing feature points and non-feature points, updating point cloud data based on the previous steps, and generating a ship scene model based on a Possion curved surface reconstruction method. The three-dimensional reconstruction method for ship virtual teaching modeling is mainly used in a point cloud data-based three-dimensional reconstruction modeling process of large equipment such as ships and the like, and is used for solving the problems of point location offset, line direction errors and the like in a point cloud generation process of various equipment in the ships and optimizing distribution of point cloud data in a reconstruction process; by reducing the density of low-feature-quantity point cloud data, reducing the total quantity of the point cloud data, reducing the load on system resources, improving the proportion of high-feature-quantity point cloud, optimizing the representation capability of the point cloud data and improving the display performance of a reconstructed picture.

Description

Technical Field

[0001] The invention belongs to the technical field of virtual teaching modeling methods, and in particular relates to a three-dimensional reconstruction method for virtual teaching modeling of ships. Background Art

[0002] As a new teaching method, scene re-enactment and simulation technology based on virtual technologies and equipment such as VR and AR can solve the problem that large equipment such as ships are not convenient for physical teaching or simulation. Through three-dimensional modeling and scene reproduction, it can achieve better teaching effects than traditional teaching methods in a smaller teaching scene and at a lower teaching cost. In order to achieve the true reproduction of real physical equipment, structure and environment, it is necessary to convert real physical information into visible data in a virtual scene through mathematical modeling, three-dimensional reconstruction and other methods. Among them, three-dimensional reconstruction uses laser scanning and other technologies to obtain point cloud information on the surface of the structure, and reconstructs the surface, lines and other structures through a series of means, which effectively improves the efficiency and display effect of reconstruction. In the process of three-dimensional reconstruction of ships and other equipment, due to the large number of equipment in the relatively closed space of the ship, more serious problems such as point cloud offset and structural line deformation occurred during the three-dimensional reconstruction process than in other occasions. Summary of the invention

[0003] The purpose of the present invention is to provide a method for improving the quality of three-dimensional reconstruction of ship teaching scenes and suppressing the point offset phenomenon during the reconstruction process based on the current situation and actual needs. The method can improve the matching degree between the reconstructed point cloud data and the actual physical structure and be used for the three-dimensional reconstruction of ship virtual teaching modeling.

[0004] To achieve the above purpose, the present invention adopts the following technical solution.

[0005] A three-dimensional reconstruction method for ship virtual teaching modeling includes the following steps:

[0006] Step 1: Establish the point cloud data coordinate system: Obtain the ship scene point cloud dataset based on point cloud scanning {p k}, where p k Represents the point cloud data obtained by the kth scan. For the point cloud data p k The i-th point p in k,i , define its coordinates in the scanning station coordinate system L as L k,i , the coordinates in the earth coordinate system W are W k,i ;

[0007] Step 2: Optimization of ship spatial contour feature points

[0008] Obtain the point cloud data of the ship scene, and establish the surface point cloud data sets F:F based on the continuous smooth surfaces in the point cloud data. i(x, y, z) = 0; for the i-th smooth surface F in the surface point cloud dataset i :F i The mth (m≤M) continuously differentiable point P on (x,y,z)=0 m :f i (x m ,y m ,z m )=0, solve for point P m The K-domain point set N pm :N pm ={p j (x j ,y j ,z j )},j=1,2....J;

[0009] Calculate point P m The K-domain point set {p k}, k=1,2....The center coordinates {x0,y0,z0} in K:

[0010]

[0011] Further, the covariance matrix C of the neighboring points in the K domain point set can be expressed as:

[0012]

[0013] The covariance matrix C is a semi-positive symmetric matrix, and its minimum eigenvalue λ is solved min and the corresponding eigenvector v min , on the smooth surface F i The minimum eigenvalue λ min and its corresponding eigenvector v min That is the smooth surface F i At point P m Normal vector of point P; Based on the above steps, determine m The normal vector of each point in the K-domain point set is set, and the normal vector angle threshold α is specified according to the simplified ratio min ; Take the normal vector corresponding to the center point {x0, y0, z0} of the K domain point set as the initial normal vector n0, and k Its nearest neighbor point p k-1 The normal vector angle α k , keep α k ≥a min Point p k Form the characteristic points of the ship's spatial contour and delete other points;

[0014] Step 3: Optimization of feature points and non-feature points

[0015] According to the aforementioned covariance matrix C, point P containing point distance weight coefficient is established m The K-domain covariance matrix C′ m : where r m Point P m The radius of the K domain; solve the eigenvalue λ′ of the covariance matrix C′ m1 ,λ′ m2 ,λ′ m3 ,λ′ m1 ≤λ′ m2 ≤λ′ m3 ; Then the rate of change of the normal vector corresponding to each point in the K domain on the surface is expressed as the eigenvalue λ′ m1 ,λ′ m2 ,λ′ m3 To ensure that the surface changes of the three-dimensional ellipsoid surface composed of basic parameters are consistent, a matching threshold is constructed Then for point P m , if ε′ m ≥ε′0, then point P m is a feature point, otherwise it is a non-feature point;

[0016] Non-feature point cloud data constitutes the smooth part of the 3D surface. By reducing the point cloud density of the smooth part, the total amount of point cloud data of the entire 3D model can be effectively reduced without affecting the key features such as the 3D surface contour, thereby improving data processing performance. The specific processing steps are as follows:

[0017] Based on the above steps, the non-feature point cloud dataset {(x′ i ,y′ i ,z′ i )}, extract the maximum value x′ of the three coordinates i,max ,y′ i,max ,z′ i,max And the minimum value of the three coordinates x′ i,min ,y′ i,min ,z′ i,min Establish a three-dimensional bounding box ΔL, the side lengths of the three-dimensional bounding box ΔL are Where Δ i >0 (i=x,y,z) means to ensure the side length margin of all points;

[0018] Based on the voxel bounding method, the three-dimensional bounding box ΔL is divided into N small bounding boxes Δl n (n=1,2...N), each cubic bounding box Δl n The side length is l i =n -1 L i (i=x,y,z); determine the bounding box Δl of each cube respectively nThe center point O n Coordinates

[0019] Where (x n,1 ,y n,1 ,z n,1 ) and (x n,2 ,y n,2 ,z n,2 ) are the cubic bounding boxes Δl n The coordinates of the two end points of any diagonal line; calculate the bounding box Δl of each cube respectively n Each point p n (x n ,y n ,z n ) and the center point O n Distance Extract the point data with the smallest distance and keep it, while deleting each cube bounding box Δl n The above segmentation process is repeated until the preset requirements are met;

[0020] Step 4: Update the point cloud data based on the above steps, and obtain valid point cloud data after necessary preprocessing; the preprocessing includes determining whether the average Euclidean distance from the point to the midpoint of the neighborhood satisfies the Gaussian distribution after obtaining the neighborhood through K nearest neighbor search for possible outliers, and removing them if not.

[0021] Step 5: Update the point cloud data and generate a ship scene model based on the Possion surface reconstruction method.

[0022] For the further improvement or preferred implementation scheme of the above-mentioned three-dimensional reconstruction method for ship virtual teaching modeling, the step 3 also includes the content of point cloud direction adjustment optimization, which specifically refers to: determining the coordinates of each point and calculating the normal direction according to the point cloud data determined in the above-mentioned step, and establishing a query sequence {p′0, p′1...p′ n ...}; Convert the normal vector of each point After translation, the normal vector after translation can be expressed as The normal vector of the first point p′0 in the query sequence is (0,0,1); the weight value of the specified point p′0 is 1, and the weight values ​​of other points are updated in the order of the query sequence. in Refers to the normal vector of the point corresponding to the minimum weight value among the updated points; in each update process, if Then direction to convert.

[0023] The further improvement or preferred implementation scheme of the above three-dimensional reconstruction method for ship virtual teaching modeling also includes a step for establishing a point cloud octree Tree and its position index and gradient, specifically: obtaining all point cloud data P all ={x i ,y i ,z i , i=1,2,3......n}, a point cloud octree Tree is established based on the aforementioned bounding box structure; each child node T at each level in the point cloud octree Tree i Corresponding to each group of bounding boxes of the same size; assuming that the established point cloud octree Tree is N layers, the side length of the minimum bounding box is l min =L max ×2 -N ;

[0024] Assume that the point cloud data P all The position index corresponding to the octree child node is q m , then the description function used to represent the corresponding child node can be expressed as:

[0025]

[0026] Where m represents the point cloud data P all The number of layers corresponding to the octree child node; q mid Represents point cloud data P all The octree child node corresponds to the midpoint of the bounding box; i = x, y, z; * indicates convolution operation; then the child node q m Indicates the gradient of the function t m Interpolate weights for child nodes.

[0027] Its beneficial effects are:

[0028] The three-dimensional reconstruction method for virtual teaching modeling of ships in the present application is mainly used in the three-dimensional reconstruction modeling process of ships and other large equipment based on point cloud data. It is used to improve the problems of point offset, line direction error, etc. in the point cloud generation process of various equipment in the ship, and optimize the distribution of point cloud data in the reconstruction process; by reducing the density of low-feature point cloud data, reducing the total amount of point cloud data, reducing the load on system resources, and increasing the proportion of high-feature point clouds, the expression ability of point cloud data is optimized, and the visualization performance of the reconstructed picture is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a schematic diagram of the voxel bounding method sampling principle. DETAILED DESCRIPTION

[0030] The present invention is described in detail below in conjunction with specific embodiments.

[0031] The present invention is mainly used to provide a method for three-dimensional modeling and reconstruction of various types of equipment in a confined space in scenarios such as virtual teaching and training, so as to realize virtual reproduction of real scenes, especially a three-dimensional reconstruction method for internal modeling of large equipment such as various ships.

[0032] Step 1: Establish the point cloud data coordinate system: Obtain the ship scene point cloud dataset based on point cloud scanning {p k}, where p k Represents the point cloud data obtained by the kth scan. For the point cloud data p k The i-th point p in k,i , define its coordinates in the scanning station coordinate system L as L k,i , the coordinates in the earth coordinate system W are W k,i ;

[0033] Point cloud data is the data bridge between the actual physical structure and the virtual three-dimensional model. The coordinates and relative position relationship of each point in the point cloud data are necessary conditions for reconstructing the physical structure. To this end, it is necessary to establish a corresponding coordinate system and generate original point cloud data through special equipment or instruments such as laser scanners and depth cameras.

[0034] In visual information, the edge contour of the surface of the object structure is the core characteristic attribute of the structure performance, and it is also an important content in the point cloud data processing process. However, since the point cloud data is discrete data, the contour edge performance is unclear, resulting in display abnormalities and other problems. For this reason, the following steps are set in this application for optimization:

[0035] Step 2: Optimization of ship spatial contour feature points

[0036] Obtain the point cloud data of the ship scene, and establish the surface point cloud data sets F:F based on the continuous smooth surfaces in the point cloud data. i (x,y,z)=0;

[0037] For the i-th smooth surface F in the surface point cloud dataset i :F i The mth (m≤M) continuously differentiable point P on (x,y,z)=0 m :f i (x m ,y m ,z m )=0, solve for point P m The K-domain point set N pm :

[0038] N pm ={p j (x j ,yj ,z j )},j=1,2....J;

[0039] Calculate point P m The K-domain point set {p k}, k=1,2....The center coordinates {x0,y0,z0} in K:

[0040]

[0041] Further, the covariance matrix C of the neighboring points in the K domain point set can be expressed as:

[0042]

[0043] The covariance matrix C is a semi-positive symmetric matrix, and its minimum eigenvalue λ can be solved min and the corresponding eigenvector v min , on the smooth surface F i The minimum eigenvalue λ min and its corresponding eigenvector v min That is the smooth surface F i At point P m Normal vector of point P; Based on the above steps, determine m The normal vector of each point in the K-domain point set is set, and the normal vector angle threshold α is specified according to the simplified ratio min ; Take the normal vector corresponding to the center point {x0, y0, z0} of the K domain point set as the initial normal vector n0, and k Its nearest neighbor point p k-1 The normal vector angle α k , keep α k ≥α min Point p k Form the characteristic points of the ship's spatial contour and delete other points;

[0044] Step 3: Optimization of feature points and non-feature points

[0045] According to the aforementioned covariance matrix C, point P containing point distance weight coefficient is established m The K-domain covariance matrix C′ m : where r m Point P m The radius of the K field;

[0046] Solve for the eigenvalue λ′ of the covariance matrix C′ m1 ,λ′ m2 ,λ′ m3 ,λ′ m1 ≤λ m2 ≤λm3 ; Then the rate of change of the normal vector corresponding to each point in the K domain on the surface is expressed as the eigenvalue λ′ m1 ,λ′ m2 ,λ′ m3 To ensure that the surface changes of the three-dimensional ellipsoid surface composed of basic parameters are consistent, a matching threshold is constructed Then for point P m , if ε′ m ≥ε′0, then point P m is a feature point, otherwise it is a non-feature point;

[0047] Non-feature point cloud data constitutes the smooth part of the 3D surface. By reducing the point cloud density of the smooth part, the total amount of point cloud data of the entire 3D model can be effectively reduced without affecting the key features such as the 3D surface contour, thereby improving data processing performance. The specific processing steps are as follows:

[0048] Based on the above steps, the non-feature point cloud dataset {(x′ i ,y′ i ,z′ i )}, extract the maximum value x′ of the three coordinates i,max ,y′ i,max ,z′ i,max And the minimum value of the three coordinates x′ i,min ,y′ i,min ,z′ i,min Establish a three-dimensional bounding box ΔL, the side lengths of the three-dimensional bounding box ΔL are Where Δ i >0 (i=x,y,z) means to ensure the side length margin of all points;

[0049] like Figure 1 As shown, based on the voxel bounding method, the three-dimensional bounding box ΔL is divided into N small bounding boxes Δl n (n=1,2...N), each cubic bounding box Δl n The side length is l i =n -1 L i (i=x,y,z);

[0050] Determine the bounding box Δl of each cube separately n The center point O n Coordinates

[0051] Where (x n,1 ,y n,1 ,z n,1 ) and (x n,2 ,y n,2 ,z n,2 ) are the cubic bounding boxes Δl nThe coordinates of the two end points of any diagonal line;

[0052] Calculate each cube bounding box Δl separately n Each point p n (x n ,y n ,z n ) and the center point O n Distance Extract the point data with the smallest distance and keep it, while deleting each cube bounding box Δl n Other points in

[0053] The above segmentation process is executed repeatedly until the preset requirements are met;

[0054] In particular, the present application also provides a step for optimizing the direction of the point cloud vector and adjusting the curve expression capability. Specifically, the coordinates of each point are determined and the normal direction is calculated according to the point cloud data determined in the above steps, and a query sequence {p′0, p′1...p′ is established in the order of height values ​​from small to large. n ...};

[0055] Convert the normal vector of each point After translation, the normal vector after translation can be expressed as Then the normal vector of the first point p′0 in the query sequence is (0,0,1);

[0056] The weight value of the designated point p′0 is 1, and the weight values ​​of other points are updated in the order of the query sequence. in It refers to the normal vector of the point corresponding to the minimum weight value among the updated points;

[0057] During each update, if Then direction to convert;

[0058] It also includes steps for establishing the point cloud octree Tree and its position index and gradient, specifically:

[0059] Get all point cloud data P all ={x i ,y i ,z i , i=1,2,3......n}, a point cloud octree Tree is established based on the aforementioned bounding box structure; each child node T at each level in the point cloud octree Tree i Corresponding to each group of bounding boxes of the same size; assuming that the established point cloud octree Tree is N layers, the side length of the minimum bounding box is l min =L max ×2 -N ;

[0060] Assume that the point cloud data P all The position index corresponding to the octree child node is q m , then the description function used to represent the corresponding child node can be expressed as:

[0061]

[0062] Where m represents the point cloud data P all The number of layers corresponding to the octree child node; q mid Represents point cloud data P all The octree child node corresponds to the midpoint of the bounding box; i = x, y, z; * indicates convolution operation;

[0063] Then the child node q m Indicates the gradient of the function t m Interpolate weights for child nodes;

[0064] Step 4: Update the point cloud data based on the above steps, and obtain valid point cloud data after necessary preprocessing; the preprocessing includes determining whether the average Euclidean distance from the point to the midpoint of the neighborhood satisfies the Gaussian distribution after obtaining the neighborhood through K nearest neighbor search for possible outliers, and removing them if not.

[0065] Step 5: Update the point cloud data and generate a ship scene model based on the Possion surface reconstruction method.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A three-dimensional reconstruction method for ship virtual teaching modeling, characterized in that: The steps include: Step 1: Establish the point cloud data coordinate system: Obtain the ship scene point cloud dataset based on point cloud scanning {p k }, where p k Represents the point cloud data obtained by the kth scan. For the point cloud data p k The i-th point p in k,i , define its coordinates in the scanning station coordinate system L as L k,i , the coordinates in the earth coordinate system W are W k,i ; Step 2: Optimization of ship spatial contour feature points Obtain the point cloud data of the ship scene, and establish the surface point cloud data sets F:F based on the continuous smooth surfaces in the point cloud data. i (x, y, z) = 0; for the i-th smooth surface F in the surface point cloud dataset i :F i The m(m￡M)th continuously differentiable point P on (x,y,z)=0 m :f i (x m ,y m ,z m )=0, solve for point P m The K-domain point set N pm :N pm ={p j (x j ,y j ,z j )},j=1,2....J; Calculate point P m The K-domain point set {p k }, k=1,2....The center coordinates {x0,y0,z0} in K: Further, the covariance matrix C of the neighboring points in the K domain point set can be expressed as: The covariance matrix C is a semi-positive symmetric matrix, and its minimum eigenvalue λ is solved min and the corresponding eigenvector v min , on the smooth surface F i The minimum eigenvalue λ min and its corresponding eigenvector v min That is the smooth surface F i At point P m Normal vector of point P; Based on the above steps, determine m The normal vector of each point in the K-domain point set is set, and the normal vector angle threshold α is specified according to the simplified ratio min ; Take the normal vector corresponding to the center point {x0, y0, z0} of the K domain point set as the initial normal vector n0, and k Its nearest neighbor point p k-1 The normal vector angle α k , keep α k 3a min Point p k Form the characteristic points of the ship's spatial contour and delete other points; Step 3: Optimization of feature points and non-feature points According to the aforementioned covariance matrix C, point P containing point distance weight coefficient is established m The K-domain covariance matrix C′ m : where r m Point P m The radius of the K domain; solve the eigenvalue λ′ of the covariance matrix C′ m1 ,λ′ m2 ,λ′ m3 ,λ′ m1 ≤λ′ m2 ≤λ′ m3 ; Then the rate of change of the normal vector corresponding to each point in the K domain on the surface is expressed as the eigenvalue λ′ m1 ,λ′ m2 ,λ′ m3 To ensure that the surface changes of the three-dimensional ellipsoid surface composed of basic parameters are consistent, a matching threshold is constructed Then for point P m , if ε′ m 3ε0′, then point P m is a feature point, otherwise it is a non-feature point; Non-feature point cloud data constitutes the smooth part of the 3D surface. By reducing the point cloud density of the smooth part, the total amount of point cloud data of the entire 3D model can be effectively reduced without affecting the key features such as the 3D surface contour, thereby improving data processing performance. The specific processing steps are as follows: Based on the above steps, the non-feature point cloud dataset {(x i ′,y i ′,z i ′)}, extract the maximum value x of the three coordinates i ' ,max ,y i ' ,max ,z i ' ,max And the minimum value of the three coordinates x i ' ,min ,y i ' ,min ,z i ' ,min Establish a three-dimensional bounding box ΔL, the side lengths of the three-dimensional bounding box ΔL are Where Δ i >0 (i=x,y,z) means to ensure the side length margin of all points; Based on the voxel bounding method, the three-dimensional bounding box ΔL is divided into N small bounding boxes Δl n (n=1,2...N), each cubic bounding box Δl n The side length is l i =n -1 L i (i=x,y,z); determine the bounding box Δl of each cube respectively n The center point O n Coordinates Where (x n,1 ,y n,1 ,z n,1 ) and (x n,2 ,y n,2 ,z n,2 ) are the cubic bounding boxes Δl n The coordinates of the two end points of any diagonal line; calculate the bounding box Δl of each cube respectively n Each point p n (x n ,y n ,z n ) and the center point O n Distance Extract the point data with the smallest distance and keep it, while deleting each cube bounding box Δl n The above segmentation process is repeated until the preset requirements are met; Step 4: Update the point cloud data based on the above steps, and obtain valid point cloud data after necessary preprocessing; the preprocessing includes determining whether the average Euclidean distance from the point to the midpoint of the neighborhood satisfies the Gaussian distribution after obtaining the neighborhood through K nearest neighbor search for possible outliers, and removing them if not. Step 5: Update the point cloud data and generate a ship scene model based on the Possion surface reconstruction method.

2. The three-dimensional reconstruction method for ship virtual teaching modeling according to claim 1 is characterized in that: The step 3 also includes the content of point cloud direction adjustment optimization, which specifically refers to: determining the coordinates of each point and calculating the normal direction according to the point cloud data determined in the previous step, and establishing a query sequence {p0′, p1′...p n ′...}; transform the normal vector of each point After translation, the normal vector after translation can be expressed as The normal vector of the first point p′0 in the query sequence is (0,0,1); the weight value of the specified point p′0 is 1, and the weight values ​​of other points are updated in the order of the query sequence. in Refers to the normal vector of the point corresponding to the minimum weight value among the updated points; in each update process, if Then direction to convert.

3. The three-dimensional reconstruction method for ship virtual teaching modeling according to claim 2 is characterized in that: It also includes steps for establishing a point cloud octree Tree and its position index and gradient, specifically: obtaining all point cloud data P all ={x i ,y i ,z i , i=1,2,3......n}, a point cloud octree Tree is established based on the aforementioned bounding box structure; each child node T at each level in the point cloud octree Tree i Corresponding to each group of bounding boxes of the same size; assuming that the established point cloud octree Tree is N layers, the side length of the minimum bounding box is l min =L max ×2 -N ; Assume that the point cloud data P all The position index corresponding to the octree child node is q m , then the description function used to represent the corresponding child node can be expressed as: Where m represents the point cloud data P all The number of layers corresponding to the octree child node; q mid Represents point cloud data P all The octree child node corresponds to the midpoint of the bounding box; i = x, y, z; * indicates convolution operation; then the child node q m Indicates the gradient of the function t m Interpolate weights for child nodes.

Images