Feature-based oil and gas field scene point cloud registration method and device and electronic equipment

Abstract

The invention provides a feature-based oil and gas field scene point cloud registration method and device and electronic equipment, and belongs to the technical field of point cloud registration. The method comprises the steps of obtaining a plurality of associated point pairs based on a first point cloud feature vector and a second point cloud feature vector; based on the first coordinates and the second coordinates, determining a target consistency matrix, the first coordinates being coordinates of points in the first point cloud in each associated point pair, and the second coordinates being coordinates of points in the second point cloud in each associated point pair; constructing a consistent set corresponding to each seed pair based on the plurality of seed pairs and the association point pairs respectively corresponding to the plurality of seed pairs; hypothesis testing is carried out on the transformation matrix corresponding to each consistent set, and a target transformation matrix is determined; and based on the target transformation matrix, registering the first point cloud and the second point cloud to obtain registered point cloud data. The feature-based oil and gas field scene point cloud registration method provided by the invention can improve the registration precision of the point cloud data of the oil and gas field scene.

Description

Technical Field

[0001] The present invention relates to the technical field of point cloud registration, and in particular to a feature-based oil and gas field scene point cloud registration method, device and electronic equipment. Background Art

[0002] In order to improve the efficiency of safe production in the petrochemical industry, digital transformation of the petrochemical industry has become a key approach. 3D reconstruction technology is a key technology in the digital transformation process. In the process of 3D reconstruction, point cloud registration technology is very important.

[0003] In the oil and gas field scenes of petrochemical industry, there are many transportation pipelines, valves and oil production equipment (such as pumping units). However, the three-dimensional point clouds on the transportation pipelines, valves and oil production equipment are often sparse and of low accuracy, which brings great challenges to point cloud registration.

[0004] The oil and gas field scene is an outdoor scene. The point cloud registration methods currently used in the oil and gas field scene are mainly based on the iterative closest point (ICP) method and the feature-based method. However, the above methods are dependent on the initial values. When the initial values ​​of the two point clouds are far apart, the point cloud registration accuracy is low and good results are often not achieved. Summary of the invention

[0005] The present invention provides a feature-based oil and gas field scene point cloud registration method, device and electronic equipment, which are used to solve the problem of low point cloud registration accuracy on transportation pipelines, valves and oil production equipment in the prior art.

[0006] The present invention provides a feature-based oil and gas field scene point cloud registration method, comprising: Based on the first point cloud feature vector and the second point cloud feature vector, a plurality of associated point pairs are obtained; the associated point pairs are any point in the first point cloud and the nearest neighbor point of the any point in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; Determine a target consistency matrix based on first coordinates and second coordinates, wherein the first coordinates are coordinates of points in each of the associated point pairs in the first point cloud, and the second coordinates are coordinates of points in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements whose values ​​in the target consistency matrix exceed a first threshold value; Based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, constructing a consistent set corresponding to each of the seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; Performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine a target transformation matrix; Based on the target transformation matrix, the first point cloud and the second point cloud are registered to obtain registered point cloud data.

[0007] In some embodiments, determining the target consistency matrix based on the first coordinate and the second coordinate includes: constructing a spatial consistency matrix based on the first coordinate and the second coordinate; determining a vector product of two of said spatial consistency matrices; The target consistency matrix is ​​determined by the dot product of the spatial consistency matrix and the vector product.

[0008] In some embodiments, before performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine the target transformation matrix, the method further includes: Based on the target consistency matrix and the feature similarity matrix, a correlation matrix is ​​constructed, wherein the feature similarity matrix is ​​determined based on feature vectors corresponding to any two pairs of correlation points in the consistent set; Performing eigendecomposition on the correlation matrix to obtain an eigenvector corresponding to a main direction, wherein the main direction is used to indicate a direction corresponding to a maximum eigenvalue after eigenvalue decomposition; Based on the eigenvector corresponding to the main direction and the coordinates of the associated point pairs in the consistent set, construct a least squares problem corresponding to the transformation matrix corresponding to the consistent set; The least squares problem is solved to obtain a transformation matrix corresponding to each consistent set.

[0009] In some embodiments, performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine the target transformation matrix includes: Performing a hypothesis test on the transformation matrix corresponding to each of the consistent sets to obtain a transformation matrix that maximizes the number of points in matching pairs, wherein the matching pairs are used to indicate correctly matched point pairs in the first point cloud and the second point cloud; Determine the associated point pairs corresponding to the inner points of all matching pairs based on the transformation matrix that maximizes the number of inner points of the matching pairs; The target transformation matrix is ​​determined based on the associated point pairs corresponding to the points in all the matching pairs.

[0010] In some embodiments, obtaining a plurality of associated point pairs based on the first point cloud feature vector and the second point cloud feature vector includes: calculate , and the dimension is A vector of is the first point cloud feature vector The k The characteristics of a point, is the second point cloud feature vector; Based on the dimensions The minimum value of the vector determines the second point cloud The m Points With the first point cloud The k Points is the associated point pair.

[0011] In some embodiments, the association matrix is ​​represented as: ; in, is the target consistency matrix, is the feature similarity matrix; the feature similarity matrix is ​​expressed as: ; in, is the first i The eigenvectors corresponding to the associated point pairs are: is the first j The eigenvectors corresponding to the associated point pairs.

[0012] In some embodiments, the transformation matrix corresponding to the consistent set is: ; in, R is the rotation matrix, t is the translation vector; is the first i A point in the first point cloud in a pair of associated points, y i for x i The nearest neighbor point in the second point cloud, is the eigenvector corresponding to the main direction, n +1 is the number of associated point pairs in the consistent set.

[0013] The present invention also provides a feature-based oil and gas field scene point cloud registration device, comprising: A matching module, used for obtaining a plurality of associated point pairs based on a first point cloud feature vector and a second point cloud feature vector; the associated point pair is a point in the first point cloud and its nearest neighbor in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used for indicating a fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used for indicating an FPFH descriptor corresponding to each point in the second point cloud; A determination module, configured to determine a target consistency matrix based on a first coordinate and a second coordinate, wherein the first coordinate is the coordinate of a point in each of the associated point pairs in the first point cloud, and the second coordinate is the coordinate of a point in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements in the target consistency matrix whose values ​​exceed a first threshold value; A construction module, configured to construct a consistent set corresponding to each seed pair based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; A test module, used for performing hypothesis test on the transformation matrix corresponding to each of the consistent sets to determine the target transformation matrix; A registration module is used to register the first point cloud and the second point cloud based on the target transformation matrix to obtain registered point cloud data.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the feature-based oil and gas field scene point cloud registration method as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for point cloud registration of oil and gas field scenes based on features as described in any one of the above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for point cloud registration of oil and gas field scenes based on features as described in any one of the above is implemented.

[0017] The feature-based oil and gas field scene point cloud registration method, device and electronic device provided by the present invention can improve the registration accuracy of point cloud data in oil and gas field scenes. Due to the importance of point cloud registration in three-dimensional reconstruction and mobile robots and the needs of oil and gas field scenes, the invention has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 It is a flow chart of the feature-based oil and gas field scene point cloud registration method provided by the present invention; Figure 2 It is one of the point cloud registration schematic diagrams of the feature-based oil and gas field scene point cloud registration method provided by the present invention; Figure 3 This is the second point cloud registration schematic diagram of the feature-based oil and gas field scene point cloud registration method provided by the present invention; Figure 4 This is the third point cloud registration schematic diagram of the feature-based oil and gas field scene point cloud registration method provided by the present invention; Figure 5 It is a structural schematic diagram of a feature-based oil and gas field scene point cloud registration system provided by the present invention; Figure 6 It is a structural schematic diagram of a feature-based oil and gas field scene point cloud registration device provided by the present invention; Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Among the related technologies, point cloud registration is an important technology in point cloud processing and analysis. Its purpose is to find the transformation relationship between two partially overlapping and unknown associated point clouds. This technology is one of the core technologies in Simultaneous Localization and Mapping (SLAM) and 3D reconstruction. Commonly used point cloud registration methods include ICP-based methods, probability-based methods, and feature-based methods.

[0022] The ICP-based method solves the transformation matrix between two point clouds by searching for the nearest point in one point cloud in another point cloud. This type of method is widely used in a variety of indoor and outdoor scenes, but it is heavily dependent on the initial value, and good results can only be achieved when the initial value is close to the true value. The probability-based method transforms the point cloud registration problem into a probability density estimation problem. This type of method requires the establishment of an optimization function for optimization and solution. Therefore, this type of method still relies on the initial value, and will fall into a local minimum if the initial value is not good. The feature-based method does not rely on the initial value at all, but the accuracy in challenging scenes needs to be improved.

[0023] The feature-based oil and gas field scene point cloud registration method provided by the present invention is suitable for point cloud registration of outdoor oil and gas field scenes.

[0024] Combine the following Figure 1-Figure 7 The invention describes a feature-based oil and gas field scene point cloud registration method, device and electronic equipment.

[0025] Figure 1 It is a flow chart of the feature-based oil and gas field scene point cloud registration method provided by the present invention.

[0026] It should be noted that the execution subject of the feature-based oil and gas field scene point cloud registration method provided by the present invention can be an electronic device, a component in an electronic device, an integrated circuit, or a chip. The electronic device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a personal digital assistant (PDA), etc., and the non-mobile electronic device can be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, etc., which is not specifically limited by the present invention.

[0027] The technical solution of the present invention is described in detail below by taking a computer executing the feature-based oil and gas field scene point cloud registration method provided by the present invention as an example.

[0028] Reference Figure 1 The feature-based oil and gas field scene point cloud registration method provided by the present invention includes: Step 110, based on the first point cloud feature vector and the second point cloud feature vector, obtain multiple associated point pairs; the associated point pair is any point in the first point cloud and the nearest neighbor point of any point in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud.

[0029] In step 110, point cloud data of the same oil and gas field in different coordinate systems are acquired to obtain a first point cloud and a second point cloud.

[0030] like Figure 2 As shown, Figure 2 The point cloud on the left in the figure represents the first point cloud. Figure 2 The point cloud on the right in the middle represents the second point cloud, and the first point cloud and the second point cloud are the two point clouds to be registered.

[0031] For the first point cloud and the second point cloud Perform feature extraction to obtain the first point cloud The point-by-point correspondence between the Fast point feature histograms (FPFH) descriptor and the second point cloud The FPFH descriptor corresponding to each point in the feature extraction method is to generate the local descriptor of a point through the normal vector of the points around the point. This method is widely used in outdoor scenes.

[0032] For the first point cloud and the second point cloud By performing feature extraction, we can get the first point cloud feature vector and the second point cloud feature vector .in, D is the dimension of the feature vector. Among them, the first point cloud feature vector It can represent the FPFH descriptor corresponding to each point in the first point cloud, and the feature vector of the second point cloud The FPFH descriptor corresponding to each point in the second point cloud can be represented.

[0033] In some embodiments, step 110 may include: calculate , and the dimension is A vector of is the first point cloud feature vector The k The characteristics of a point, is the second point cloud feature vector; Based on the dimension The minimum value of the vector of m Points The first point cloud k Points For the associated point pair.

[0034] In actual implementation, after completing the feature extraction of the first point cloud and the second point cloud, the feature vector of the first point cloud is and the second point cloud feature vector Perform feature matching. The specific steps of feature matching are as follows: For the first point cloud feature vector The features of any point in , find its eigenvector in the second point cloud The nearest neighbor point in . The nearest neighbor point refers to the point with the smallest Euclidean distance or the point with a Euclidean distance less than a preset threshold.

[0035] The specific calculation method of the associated point pair is as follows: (1) Obtain , and obtain a dimension of The feature vector of .

[0036] (2) Find the subscript corresponding to the minimum value of the vector m , then the first point cloud The k Points and the second point cloud The m Points Form a set of associated point pairs.

[0037] For the first point cloud Each point and its corresponding feature in the second point cloud After feature matching, multiple associated point pairs can be obtained.

[0038] Step 120, based on the first coordinate and the second coordinate, determine the target consistency matrix, the first coordinate is the coordinate of the point in the first point cloud of each associated point pair, and the second coordinate is the coordinate of the point in the second point cloud of each associated point pair; the target consistency matrix is ​​used to select multiple seed pairs, and the seed pairs are the associated point pairs corresponding to the elements whose values ​​in the target consistency matrix exceed the first threshold.

[0039] In step 120, it is assumed that N groups of associated point pairs are obtained through feature extraction and feature matching, and the set composed of these associated point pairs is recorded as . Associated point pair set Each element in , , ,and and is the key point coordinates of the matched associated point pair, for The corresponding matching points after transformation. is the first coordinate, is the second coordinate.

[0040] According to the first coordinate and the second coordinate , a target consistency matrix can be constructed.

[0041] In some embodiments, step 120 may include: Based on the first coordinate and the second coordinate, a spatial consistency matrix is ​​constructed; Determine the vector product of two spatial consistency matrices; The scalar product of the spatial consistency matrix and the vector product is determined as the target consistency matrix.

[0042] In actual implementation, according to the collection The spatial consistency matrix is ​​constructed using formula (1) .

[0043] (1)

[0044] (2)

[0045] in, d ( , )= , d ( , )= , Indicates the distance threshold.

[0046] It should be noted that for a set of correct associated point pairs, the structural relationship between adjacent points in the local range should be the same. , the feature points in the first point cloud are , there are other feature points in its neighborhood . Feature points With other feature points The distance between d ( , ), then in the associated point pair, the feature point The corresponding matching points , its neighborhood should also contain other feature points Other corresponding matching points . matching point Matching points with other The distance between d ( , ).thus, d ( , )and d ( , ) should be the same. Among them, the Euclidean distance can be used for distance calculation.

[0047] Represents the feature points in the first point cloud With other feature points The distance between the corresponding matching point in the second point cloud Matching points with other The distance between them, the difference between them.

[0048] According to the constructed spatial consistency matrix Get the second-order spatial consistency matrix , which is used to select seed pairs and construct consistent sets. The target consistency matrix is ​​the second-order spatial consistency matrix. The calculation formula is shown in formula (3): (3)

[0049] in, It means the multiplication of the corresponding elements in the same position of the matrix, that is, the scalar product; Represents matrix multiplication, that is, vector product.

[0050] Spatial consistency matrix and the goal consistency matrix After the construction is completed, the seed pair is selected. The specific selection process is as follows: (1) Target consistency matrix Find the maximum value of each row of elements. The elements in the target consistency matrix are the associated point pairs.

[0051] (2) Sort each seed pair corresponding to the maximum value from large to small, and select the associated point pair with the largest maximum value of the target consistency matrix as the seed pair. In this case, a first threshold can be set, and the associated point pair with the maximum value exceeding the first threshold is determined as the seed pair.

[0052] In addition, for the uniformity of distribution, non-maximum suppression is performed to obtain the previous Seed pairs, front The seed pairs form a set .

[0053] Step 130: construct a consistent set corresponding to each seed pair based on multiple seed pairs and the associated point pairs corresponding to the multiple seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs.

[0054] In step 130, for the seed pair set Each seed pair in , in the target consistency matrix Perform K nearest neighbor search to get the corresponding n The seed pair is related to this n The associated point pairs constitute the consistent set corresponding to the seed pair .

[0055] In some embodiments, before step 140, the method further includes: Based on the target consistency matrix and the feature similarity matrix, a correlation matrix is ​​constructed. The feature similarity matrix is ​​determined based on the feature vectors corresponding to any two correlation point pairs in the consistent set; Perform eigendecomposition on the correlation matrix to obtain the eigenvector corresponding to the main direction. The main direction is used to indicate the direction corresponding to the maximum eigenvalue after eigenvalue decomposition; Based on the eigenvectors corresponding to the main directions and the coordinates of the associated point pairs in the consistent set, the least squares problem corresponding to the transformation matrix corresponding to the consistent set is constructed; The least squares problem is solved to obtain the transformation matrix corresponding to each consistent set.

[0056] In actual implementation, the consensus set constructed for each seed pair is , calculate the feature similarity matrix according to formula (4) .

[0057] (4)

[0058] in, For the consistent concentration i The eigenvectors corresponding to the associated point pairs are: The first j The eigenvectors corresponding to the associated point pairs.

[0059] In this step, the goal consistency matrix and feature similarity matrix , construct the correlation matrix .

[0060] In some embodiments, the affinity matrix is ​​represented as: (5)

[0061] in, The target consistency matrix Middle i Line j The elements corresponding to the columns, is the feature similarity matrix Middle i Line j The element corresponding to the column.

[0062] Correlation Matrix Perform eigendecomposition to obtain the eigenvector corresponding to the main direction Among them, the main direction is the direction corresponding to the largest eigenvalue after eigenvalue decomposition.

[0063] Based on the eigenvector corresponding to the main direction and the coordinates of the associated point pairs in the consistent set , construct the least squares problem corresponding to the transformation matrix corresponding to the consistent set.

[0064] In some embodiments, the transformation matrix corresponding to the consistent set is: (6)

[0065] in, R is the rotation matrix, t is the translation vector; For the consistent concentration i The point in the first point cloud of the associated point pair, y i for x i The nearest neighbor point in the second point cloud, is the eigenvector corresponding to the principal direction, n +1 is the number of associated point pairs in the consistent set.

[0066] Construct the least squares problem corresponding to formula (6).

[0067] Then, the least squares problem is solved using singular value decomposition (SVD) to obtain the transformation matrix corresponding to each consistent set. The transformation matrix can include the rotation matrix and translation vectors .for Seed pairs, corresponding to consistent set, so we can calculate Group .

[0068] Step 140: Perform hypothesis testing on the transformation matrix corresponding to each consistent set to determine the target transformation matrix.

[0069] In actual implementation, hypothesis testing is used to obtain the transformation matrix with the highest number of matched pair internal points in each consistent set, where the matched pair internal points represent the correct associated point pairs.

[0070] It should be noted that for a point and the points in its neighborhood, if the associated points are aligned and their matching points are all correct points, then these matching points tend to be clustered together, while if they are incorrect matching points, then these points tend to be scattered.

[0071] In some embodiments, step 140 includes: Performing a hypothesis test on the transformation matrix corresponding to each consistent set to obtain a transformation matrix that maximizes the number of points in matching pairs, wherein the matching pairs are used to indicate correctly matched point pairs in the first point cloud and the second point cloud; Determine the associated point pairs corresponding to the inner points of all matching pairs based on the transformation matrix that maximizes the number of inner points of the matching pairs; Based on the associated point pairs corresponding to the points in all matching pairs, the target transformation matrix is ​​determined.

[0072] In actual implementation, from the multiple groups obtained previously The maximum number of points in all matching pairs is obtained .

[0073] In this step, the calculated value in step 130 is Group , and formula (7) calculates the transformation matrix that best corresponds to most of the associated point pairs, that is, the target transformation matrix.

[0074] (7)

[0075] in, is a distance threshold.

[0076] According to the obtained Find the interior point set , the elements in the interior point set represent all the correct associated point pairs.

[0077] According to the associated point pairs corresponding to all the interior points in the interior point set, the final target transformation matrix, i.e. the rotation matrix, is recalculated using the least squares method. and translation vectors .

[0078] Step 150: Based on the target transformation matrix, the first point cloud and the second point cloud are registered to obtain registered point cloud data.

[0079] In step 150, after obtaining the target transformation matrix, the first point cloud and the second point cloud are registered to obtain the registered point cloud data, such as Figure 3 and Figure 4 shown.

[0080] Figure 3 Indicates that the second point cloud is transformed to the coordinate system of the first point cloud. Figure 4 Indicates the transformation of the first point cloud to the coordinate system of the second point cloud.

[0081] According to the registered point cloud data, the position information of the oil and gas fields corresponding to the first point cloud and the second point cloud can be obtained.

[0082] The feature-based oil and gas field scene point cloud registration method provided by the present invention can improve the registration accuracy of point cloud data in oil and gas field scenes. Due to the importance of point cloud registration in three-dimensional reconstruction and mobile robots and the needs of oil and gas field scenes, the invention has broad application prospects.

[0083] In addition, for the set of associated pairs, the features corresponding to each associated point pair are obtained according to the Nonlocal module guided by spatial consistency, and the feature consistency matrix is ​​obtained , where A is the dimension of the feature vector corresponding to the associated point pair.

[0084] For the Nonlocal module guided by spatial consistency, the present invention adopts a weighted addition method of a classification loss function and a loss function based on a feature consistency matrix, and uses the weighted addition of these two loss functions to obtain a new loss function to train the network corresponding to the entire algorithm provided by the present invention, as shown in formula (8): (8)

[0085] in, represents the classification loss function; represents the feature consistency matrix loss function,weight; Is a fixed value that can be used to adjust the weights of the two loss functions.

[0086] In the present invention, The specific representation is as follows: (9)

[0087] in, Indicates the labels corresponding to the associated point pairs of the final point cloud output by calculation. According to the true value of the pose and The obtained true value label. Represents the Binary Cross Entropy Loss Function (BCEloss) in the classification problem.

[0088] Ground truth label , and its calculation formula is as follows: (10)

[0089] In formula (8) The calculation formula is as follows: (11)

[0090] in, , the elements in the matrix are expressed as follows: (12)

[0091] like Figure 5 As shown, the present invention proposes a feature-based oil and gas scene point cloud registration system, which includes three parts: a feature extraction module, a feature matching module and an external point elimination module.

[0092] The feature extraction module is used to effectively extract local feature descriptors in the two point clouds to be registered according to the characteristics of the oil and gas field scene where there are many transportation pipelines, valves and oil production equipment.

[0093] The feature matching module is used to perform nearest neighbor matching based on the point-by-point dense feature maps obtained from two point clouds to obtain a set of associated point pairs.

[0094] The external point elimination module is used to select seed points, construct consistent sets, calculate transformation matrices and test hypotheses on the obtained set of associated point pairs to obtain matching internal points and target transformation matrices, and finally solve the aligned point cloud data to obtain the final posture information corresponding to the oil and gas field.

[0095] The calculation process of the outlier removal module is as follows: Remove the wrong association point pairs in the established association. For the association pair set obtained after feature matching, first calculate the spatial consistency and the second-order spatial consistency matrix, and select the seed pairs based on the obtained spatial consistency matrix and the second-order spatial consistency matrix. Then, for each selected seed pair, find the K nearest neighbors in the second-order spatial consistency matrix to construct the consistent set corresponding to each seed. For the constructed consistent set, calculate the transformation matrix corresponding to each consistent set; finally, obtain the final alignment result based on the hypothesis test. In addition, construct the feature vector corresponding to each seed according to a Nonlocal module, and train the network using the loss function.

[0096] The feature-based oil and gas field scene point cloud registration device provided by the present invention is described below. The feature-based oil and gas field scene point cloud registration device described below and the feature-based oil and gas field scene point cloud registration method described above can refer to each other.

[0097] Figure 6 Schematic diagram of the structure of the feature-based oil and gas field scene point cloud registration device provided by the present invention. Figure 6 The feature-based oil and gas field scene point cloud registration device provided by the present invention includes: The matching module 610 is used to obtain a plurality of associated point pairs based on the first point cloud feature vector and the second point cloud feature vector; the associated point pair is any point in the first point cloud and its nearest neighbor in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; A determination module 620 is used to determine a target consistency matrix based on a first coordinate and a second coordinate, wherein the first coordinate is the coordinate of a point in each of the associated point pairs in the first point cloud, and the second coordinate is the coordinate of a point in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements whose values ​​in the target consistency matrix exceed a first threshold value; A construction module 630 is used to construct a consistent set corresponding to each seed pair based on the multiple seed pairs and the associated point pairs corresponding to the multiple seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; A test module 640 is used to perform a hypothesis test on the transformation matrix corresponding to each of the consistent sets to determine a target transformation matrix; The registration module 650 is used to register the first point cloud and the second point cloud based on the target transformation matrix to obtain registered point cloud data.

[0098] The feature-based oil and gas field scene point cloud registration device provided by the present invention can improve the registration accuracy of point cloud data in oil and gas field scenes. Due to the importance of point cloud registration in three-dimensional reconstruction and mobile robots and the needs of oil and gas field scenes, the invention has broad application prospects.

[0099] In some embodiments, the determining module 620 is specifically configured to: constructing a spatial consistency matrix based on the first coordinate and the second coordinate; determining a vector product of two of said spatial consistency matrices; The target consistency matrix is ​​determined by the dot product of the spatial consistency matrix and the vector product.

[0100] In some embodiments, the apparatus further comprises: A transformation module, configured to perform hypothesis testing on the transformation matrix corresponding to each of the consistent sets, and to construct a correlation matrix based on the target consistency matrix and a feature similarity matrix before determining a target transformation matrix, wherein the feature similarity matrix is ​​determined based on feature vectors corresponding to any two associated point pairs in the consistent set; Performing eigendecomposition on the correlation matrix to obtain an eigenvector corresponding to a main direction, wherein the main direction is used to indicate a direction corresponding to a maximum eigenvalue after eigenvalue decomposition; Based on the eigenvector corresponding to the main direction and the coordinates of the associated point pairs in the consistent set, construct a least squares problem corresponding to the transformation matrix corresponding to the consistent set; The least squares problem is solved to obtain a transformation matrix corresponding to each consistent set.

[0101] In some embodiments, the inspection module 640 is specifically used to: Performing a hypothesis test on the transformation matrix corresponding to each of the consistent sets to obtain a transformation matrix that maximizes the number of points in matching pairs, wherein the matching pairs are used to indicate correctly matched point pairs in the first point cloud and the second point cloud; Determine the associated point pairs corresponding to the inner points of all matching pairs based on the transformation matrix that maximizes the number of inner points of the matching pairs; The target transformation matrix is ​​determined based on the associated point pairs corresponding to the points in all the matching pairs.

[0102] In some embodiments, the matching module 610 is specifically configured to: calculate , and the dimension is A vector of is the first point cloud feature vector The k The characteristics of a point, is the second point cloud feature vector; Based on the dimensions The minimum value of the vector determines the second point cloud The m Points With the first point cloud The k Points is the associated point pair.

[0103] In some embodiments, the association matrix is ​​represented as: ; in, is the target consistency matrix, is the feature similarity matrix; the feature similarity matrix is ​​expressed as: ; in, is the first i The eigenvectors corresponding to the associated point pairs are: is the first j The eigenvectors corresponding to the associated point pairs.

[0104] In some embodiments, the transformation matrix corresponding to the consistent set is: ; in, R is the rotation matrix, t is the translation vector; is the first i A point in the first point cloud in a pair of associated points, y i for x i The nearest neighbor point in the second point cloud, is the eigenvector corresponding to the main direction, n +1 is the number of associated point pairs in the consistent set.

[0105] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730 and a communication bus 740, wherein the processor 710, the communication interface 720 and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logic instructions in the memory 730 to execute the feature-based oil and gas field scene point cloud registration method, which includes: Based on the first point cloud feature vector and the second point cloud feature vector, a plurality of associated point pairs are obtained; the associated point pairs are any point in the first point cloud and the nearest neighbor point of the any point in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; Determine a target consistency matrix based on first coordinates and second coordinates, wherein the first coordinates are coordinates of points in each of the associated point pairs in the first point cloud, and the second coordinates are coordinates of points in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements whose values ​​in the target consistency matrix exceed a first threshold value; Based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, constructing a consistent set corresponding to each of the seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; Performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine a target transformation matrix; Based on the target transformation matrix, the first point cloud and the second point cloud are registered to obtain registered point cloud data.

[0106] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0107] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the feature-based oil and gas field scene point cloud registration method provided by the above methods, the method comprising: Based on the first point cloud feature vector and the second point cloud feature vector, a plurality of associated point pairs are obtained; the associated point pairs are any point in the first point cloud and the nearest neighbor point of the any point in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; Determine a target consistency matrix based on first coordinates and second coordinates, wherein the first coordinates are coordinates of points in each of the associated point pairs in the first point cloud, and the second coordinates are coordinates of points in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements whose values ​​in the target consistency matrix exceed a first threshold value; Based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, construct a consistent set corresponding to each of the seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; Performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine a target transformation matrix; Based on the target transformation matrix, the first point cloud and the second point cloud are registered to obtain registered point cloud data.

[0108] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to perform the feature-based oil and gas field scene point cloud registration method provided by the above methods, the method comprising: Based on the first point cloud feature vector and the second point cloud feature vector, a plurality of associated point pairs are obtained; the associated point pairs are any point in the first point cloud and the nearest neighbor point of the any point in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; Determine a target consistency matrix based on first coordinates and second coordinates, wherein the first coordinates are coordinates of points in each of the associated point pairs in the first point cloud, and the second coordinates are coordinates of points in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements whose values ​​in the target consistency matrix exceed a first threshold value; Based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, construct a consistent set corresponding to each of the seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; Performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine a target transformation matrix; Based on the target transformation matrix, the first point cloud and the second point cloud are registered to obtain registered point cloud data.

[0109] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0110] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A feature-based point cloud registration method for oil and gas field scenes, characterized in that: include: Based on the first point cloud feature vector and the second point cloud feature vector, a plurality of associated point pairs are obtained; the associated point pairs are any point in the first point cloud and the nearest neighbor point of the any point in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; Determining a target consistency matrix based on first coordinates and second coordinates, wherein the first coordinates are coordinates of points in each of the associated point pairs in the first point cloud, and the second coordinates are coordinates of points in each of the associated point pairs in the second point cloud; The target consistency matrix is ​​used to select a plurality of seed pairs, where the seed pairs are the association point pairs corresponding to the elements whose values ​​in the target consistency matrix exceed the first threshold; Based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, construct a consistent set corresponding to each of the seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; Performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine a target transformation matrix; Based on the target transformation matrix, the first point cloud and the second point cloud are registered to obtain registered point cloud data.

2. The feature-based point cloud registration method for oil and gas field scenes according to claim 1 is characterized in that: The step of determining a target consistency matrix based on the first coordinate and the second coordinate includes: constructing a spatial consistency matrix based on the first coordinate and the second coordinate; determining a vector product of two of said spatial consistency matrices; The target consistency matrix is ​​determined by the dot product of the spatial consistency matrix and the vector product.

3. The feature-based point cloud registration method for oil and gas field scenes according to claim 1 is characterized in that: Before performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine the target transformation matrix, the method further includes: Based on the target consistency matrix and the feature similarity matrix, a correlation matrix is ​​constructed, wherein the feature similarity matrix is ​​determined based on feature vectors corresponding to any two pairs of correlation points in the consistent set; Performing eigendecomposition on the correlation matrix to obtain an eigenvector corresponding to a main direction, wherein the main direction is used to indicate a direction corresponding to a maximum eigenvalue after eigenvalue decomposition; Based on the eigenvector corresponding to the main direction and the coordinates of the associated point pairs in the consistent set, construct a least squares problem corresponding to the transformation matrix corresponding to the consistent set; The least squares problem is solved to obtain a transformation matrix corresponding to each consistent set.

4. The feature-based point cloud registration method for oil and gas field scenes according to claim 1, characterized in that: The performing hypothesis testing on the transformation matrix corresponding to each of the consistent sets to determine the target transformation matrix includes: Performing a hypothesis test on the transformation matrix corresponding to each of the consistent sets to obtain a transformation matrix that maximizes the number of inliers of matching pairs, wherein the matching pairs are used to indicate correctly matched point pairs in the first point cloud and the second point cloud; Determine the associated point pairs corresponding to the inner points of all matching pairs based on the transformation matrix that maximizes the number of inner points of the matching pairs; The target transformation matrix is ​​determined based on the associated point pairs corresponding to the points in all the matching pairs.

5. The feature-based point cloud registration method for oil and gas field scenes according to claim 1, characterized in that: The method of obtaining a plurality of associated point pairs based on the first point cloud feature vector and the second point cloud feature vector includes: calculate , and the dimension is A vector of is the first point cloud feature vector The k The characteristics of a point, is the second point cloud feature vector; Based on the dimensions The minimum value of the vector determines the second point cloud The m Points With the first point cloud The k Points is the associated point pair.

6. The feature-based point cloud registration method for oil and gas field scenes according to claim 3 is characterized in that: The correlation matrix is ​​expressed as: ； in, is the target consistency matrix, is the feature similarity matrix; the feature similarity matrix is ​​expressed as: ； in, is the first i The eigenvectors corresponding to the associated point pairs are: is the first j The eigenvectors corresponding to the associated point pairs.

7. The feature-based point cloud registration method for oil and gas field scenes according to claim 6 is characterized in that: The transformation matrix corresponding to the consistent set is: ； in, R is the rotation matrix, t is the translation vector; is the first i A point in the first point cloud in a pair of associated points, y i for x i The nearest neighbor point in the second point cloud, is the eigenvector corresponding to the main direction, n +1 is the number of associated point pairs in the consistent set.

8. A feature-based point cloud registration device for oil and gas field scenes, characterized in that: include: A matching module, used for obtaining a plurality of associated point pairs based on the first point cloud feature vector and the second point cloud feature vector; The associated point pair is any point in the first point cloud and its nearest neighbor in the second point cloud; the first point cloud and the second point cloud are different point cloud data corresponding to the same oil and gas field, the first point cloud feature vector is used to indicate the fast point feature histogram FPFH descriptor corresponding to each point in the first point cloud, and the second point cloud feature vector is used to indicate the FPFH descriptor corresponding to each point in the second point cloud; A determination module, configured to determine a target consistency matrix based on a first coordinate and a second coordinate, wherein the first coordinate is the coordinate of a point in each of the associated point pairs in the first point cloud, and the second coordinate is the coordinate of a point in each of the associated point pairs in the second point cloud; the target consistency matrix is ​​used to select a plurality of seed pairs, wherein the seed pairs are associated point pairs corresponding to elements in the target consistency matrix whose values ​​exceed a first threshold value; A construction module, configured to construct a consistent set corresponding to each seed pair based on the multiple seed pairs and the associated point pairs respectively corresponding to the multiple seed pairs, wherein the associated point pairs corresponding to the seed pairs are obtained after performing a K-nearest neighbor search on the target consistency matrix based on the seed pairs; A test module, used for performing hypothesis test on the transformation matrix corresponding to each of the consistent sets to determine the target transformation matrix; A registration module is used to register the first point cloud and the second point cloud based on the target transformation matrix to obtain registered point cloud data.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the feature-based point cloud registration method for oil and gas field scenes as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the feature-based point cloud registration method for oil and gas field scenes as described in any one of claims 1 to 7 is implemented.

Images