A point cloud matching method and device based on quasi-newton optimization

Abstract

The application relates to the technical field of point cloud matching, in particular to a point cloud matching method and device based on quasi-Newton optimization, which can solve the problem that the maximum distance deviation between the matched actual point cloud and the corresponding points of the template point cloud is not small enough in the prior art. The point cloud matching method comprises the following steps: obtaining a template point cloud, an initial actual point cloud, an initial transformation relationship between the template point cloud and the initial actual point cloud, and a target function; performing iterative optimization on the target function by using a quasi-Newton algorithm to obtain an optimized transformation relationship, which is used for transforming the initial actual point cloud to obtain an optimized actual point cloud; updating the target function based on the optimized actual point cloud and the optimized transformation relationship to obtain a new target function, which is used as the target function in the next iteration; stopping the iteration when the target function reaches a convergence standard, and determining a target transformation relationship based on the optimized transformation relationships obtained in the previous iterations.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of point cloud matching, in particular to a point cloud matching method and device based on quasi-Newton optimization. BACKGROUND

[0002] In the field of visual images, point clouds can be used to reflect the three-dimensional geometric shape of the visible surface of an object, wherein the coordinate values of the points in the point cloud represent the positions of the points in the scene in space.

[0003] In point cloud processing, different poses of multiple point clouds are transformed into the same perspective by a point cloud matching algorithm, and multiple point clouds at the same perspective can form a complete point cloud for subsequent detection and measurement, or after the actual point cloud is matched with the template point cloud, the point cloud deviation is obtained by comparison.

[0004] The point cloud matching algorithm is a process of calculating the rigid body transformation relationship between the actual point cloud and the template point cloud. The existing point cloud matching algorithm usually uses the least square method, but the least square method cannot guarantee that the maximum distance deviation between the corresponding points of the matched actual point cloud and the template point cloud is small enough. SUMMARY

[0005] In order to solve the problem that the existing technology cannot make the maximum distance deviation between the corresponding points of the matched actual point cloud and the template point cloud small enough, the present application provides a point cloud matching method and device based on quasi-Newton optimization.

[0006] The embodiments of the present application are implemented as follows:

[0007] The embodiments of the present application provide a point cloud matching method based on quasi-Newton optimization, which comprises:

[0008] Obtaining a template point cloud, an initial actual point cloud, and an initial transformation relationship between the template point cloud and the initial actual point cloud, wherein the template point cloud, the initial actual point cloud and the initial transformation relationship are used to construct a target function;

[0009] Using a quasi-Newton algorithm to iteratively optimize the target function to obtain an optimized transformation relationship, wherein the optimized transformation relationship is used to transform the initial actual point cloud to obtain an optimized actual point cloud;

[0010] Updating the target function based on the optimized actual point cloud and the optimized transformation relationship to obtain a new target function, wherein the new target function is used as the target function in the next iteration;

[0011] stopping the iteration when the objective function reaches a convergence criterion, and determining a target transformation relationship based on the optimization transformation relationship obtained in each iteration, wherein the target transformation relationship is used to match the template point cloud and the initial actual point cloud, so that the maximum distance deviation value between the template point cloud and the initial actual point cloud after matching is smaller.

[0012] In some embodiments, the transformation relationship includes a spatial rotation angle and a three-dimensional translation amount, the spatial rotation angle is used to determine a rotation matrix between the template point cloud and the actual point cloud, and the three-dimensional translation amount is used to represent a translation vector between the template point cloud and the actual point cloud.

[0013] In some embodiments, the initial transformation relationship includes an initial spatial rotation angle and an initial three-dimensional translation amount, the template point cloud, the initial actual point cloud and the initial transformation relationship are used to construct an objective function, and further comprising:

[0014] determining an initial rotation matrix between the template point cloud and the initial actual point cloud based on the initial spatial rotation angle;

[0015] constructing a deviation equation based on the initial rotation matrix, the initial three-dimensional translation vector, the template point cloud and the initial actual point cloud, the deviation equation being used to represent the deviation between corresponding points in the template point cloud and the initial actual point cloud;

[0016] constructing the objective function based on the deviation equation and the initial transformation relationship.

[0017] In some embodiments, the objective function is iteratively optimized by using a quasi-Newton algorithm to obtain an optimization transformation relationship, and further comprising:

[0018] initializing the coefficients of the objective function and the initial transformation relationship to obtain an initial value;

[0019] substituting the initial value into the objective function to perform quasi-Newton iterative optimization, and calculating the gradient of the objective function at each iteration;

[0020] if the gradient of the objective function is less than a gradient preset value or the current iteration number is greater than a first preset iteration number, outputting the optimization transformation relationship.

[0021] In some embodiments, the optimization transformation relationship is used to transform the initial actual point cloud to obtain an optimized actual point cloud, and further comprising:

[0022] if the difference between the optimization transformation relationship and the initial transformation relationship is not within a preset range, transforming the initial actual point cloud using the optimization transformation relationship to obtain the optimized actual point cloud.

[0023] In some embodiments, the quasi-Newton iterative optimization is further comprising, after the initial value is substituted into the objective function and the gradient of the objective function at each iteration is calculated:

[0024] If the gradient of the objective function is greater than or equal to a gradient threshold or the current iteration number is less than the first preset iteration number, a search direction is calculated based on the gradient of the objective function at the current iteration and the initial matrix of the approximate Hessian matrix;

[0025] A search step is calculated based on the search direction and the optimization transformation relationship of the current iteration;

[0026] The approximate Hessian matrix is calculated based on the search direction, the search step, the optimization transformation relationship of the current iteration and the gradient of the objective function;

[0027] The optimization transformation relationship of the current iteration is updated based on the approximate Hessian matrix;

[0028] The gradient of the objective function of the next iteration is calculated based on the updated optimization transformation relationship until the gradient of the objective function is less than a gradient preset value or the current iteration number is greater than a preset iteration number.

[0029] In some embodiments, the convergence criterion is that the difference between the optimization transformation relationship obtained by the iterative optimization and the initial transformation relationship is within a preset range.

[0030] In some embodiments, the target transformation relationship is determined based on the optimization transformation relationships obtained by the previous iterations, and the method further comprises:

[0031] The optimization transformation relationships obtained by the previous iterations are accumulated and summed to obtain the target transformation relationship.

[0032] In another aspect, the application also provides a point cloud matching device based on quasi-Newton optimization, comprising:

[0033] An acquisition module is configured to acquire a template point cloud, an initial actual point cloud and an initial transformation relationship between the template point cloud and the initial actual point cloud, wherein the template point cloud, the initial actual point cloud and the initial transformation relationship are used to construct an objective function;

[0034] An iteration module is configured to perform iterative optimization on the objective function by using a quasi-Newton algorithm to obtain an optimization transformation relationship, wherein the optimization transformation relationship is used to transform the initial actual point cloud to obtain an optimized actual point cloud;

[0035] An objective function updating module is configured to update the objective function based on the optimized actual point cloud and the optimization transformation relationship to obtain a new objective function, wherein the new objective function is used as the objective function for the next iteration;

[0036] determining module configured to stop iteration when the objective function reaches a convergence criterion, and determine a target transformation relationship based on the optimization transformation relationships obtained in previous iterations, wherein the target transformation relationship is used to match the template point cloud and the initial actual point cloud, so that the maximum distance deviation between the template point cloud and the initial actual point cloud after matching is smaller.

[0037] In some embodiments, the transformation relationship includes a spatial rotation angle and a three-dimensional translation amount, the spatial rotation angle is used to determine a rotation matrix between the template point cloud and the actual point cloud, and the three-dimensional translation amount is used to represent a translation vector between the template point cloud and the actual point cloud.

[0038] In some embodiments,

[0039] Advantages of the present application: In the present application, the minimization problem of the maximum distance deviation between the actual point cloud and the template point cloud is expressed by constructing an objective function, the objective function is iteratively optimized by using a quasi-Newton algorithm, a plurality of optimization transformation relationships are output after multiple iterations, a target transformation relationship is determined based on the plurality of optimization transformation relationships, and the initial actual point cloud is converted based on the target transformation relationship, so that the matching accuracy between the converted actual point cloud and the template point cloud is higher and the maximum distance deviation is smaller. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0041] Figure 1 A flowchart of a point cloud matching method based on quasi-Newton optimization in an embodiment of the present application is shown;

[0042] Figure 2 A flowchart of constructing an objective function in another embodiment of the present application is shown;

[0043] Figure 3 A flowchart of iteratively optimizing the objective function by using a quasi-Newton algorithm to obtain an optimization transformation relationship in another embodiment of the present application is shown;

[0044] Figure 4 A flowchart of converting an initial actual point cloud to obtain an optimized actual point cloud in another embodiment of the present application is shown;

[0045] Figure 5Fig. 2 shows a flowchart of a process performed after computing the gradient of the objective function in each iteration of the quasi-Newton iterative optimization of substituting the initial value into the objective function according to another embodiment of the present application;

[0046] Figure 6 Fig. 3 shows a flowchart of a process of matching the template point cloud and the initial actual point cloud using the target transformation relationship according to another embodiment of the present application;

[0047] Figure 7 Fig. 4 shows a flowchart of a point cloud matching method based on quasi-Newton optimization according to another embodiment of the present application;

[0048] Figure 8 Fig. 5 shows a structure diagram of a point cloud matching device based on quasi-Newton optimization according to another embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, implementation and advantages of the present application more clear, the following will combine the drawings in the exemplary embodiments of the present application to clearly and completely describe the exemplary embodiments of the present application. Obviously, the described exemplary embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application.

[0050] It should be noted that the brief description of the terms in the present application is only for the convenience of understanding the following described embodiments, and is not intended to limit the embodiments of the present application. Unless otherwise specified, these terms should be understood according to their ordinary and general meanings.

[0051] The terms "first", "second", "third", etc. in the specification and claims of the present application and the above drawings are used to distinguish similar or similar objects or entities, and do not necessarily mean to limit the specific order or sequence, unless otherwise specified. It should be understood that the terms used in this way can be interchanged under appropriate circumstances.

[0052] The terms "include" and "have" and any variations thereof are intended to cover but not exclusive inclusion, for example, a product or device including a series of components does not necessarily limit to all the components clearly listed, but can include other components not clearly listed or inherent to these products or devices.

[0053] The terms "set", "connected" should be understood broadly, for example, it can be fixedly connected, or detachably connected, or integrally connected; it can be directly connected, or indirectly connected through an intermediate medium, or the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0054] Figure 1A flowchart of a point cloud matching method based on quasi-Newton optimization in an embodiment of the present application is shown. As shown in Figure 1 A point cloud matching method based on quasi-Newton optimization, the method comprising:

[0055] In step 110, a template point cloud, an initial actual point cloud, and an initial transformation relationship between the template point cloud and the initial actual point cloud are obtained, and the template point cloud, the initial actual point cloud, and the initial transformation relationship are used to construct an objective function.

[0056] A point cloud is a collection of mass points expressing a target spatial distribution and a target surface characteristic under the same spatial reference system. After obtaining the spatial coordinates of each sampling point on the surface of an object, a collection of points is obtained, which is referred to as a point cloud. With continuous breakthroughs in computer vision technology and sensor technology, methods for generating object point clouds through laser scanning have rapidly developed and improved. In this embodiment, a template point cloud and an initial actual point cloud are obtained through a depth camera, a binocular camera, or a 3D laser scanning camera.

[0057] Before the template point cloud and the initial actual point cloud are matched or when the matching accuracy between the template point cloud and the initial actual point cloud is low, the transformation coefficients therebetween are obtained as an initial transformation relationship. Since the template point cloud and the initial actual point cloud are not matched, the pose difference therebetween is large, the deviation between the initial transformation relationship and the target transformation relationship is large, and the maximum distance deviation value between the template point cloud and the initial actual point cloud is also large.

[0058] In some embodiments, the initial transformation relationship is represented by six parameters, including an initial spatial rotation angle (a, b, g) and an initial three-dimensional translation amount (T x ,T y ,T z ). The spatial rotation angle is used to determine a rotation matrix between the template point cloud and the actual point cloud, and the three-dimensional translation amount is used to represent a translation vector between the template point cloud and the actual point cloud. In the field of three-dimensional point clouds, the rigid body transformation relationship between two point clouds can be represented by a rotation matrix and a three-dimensional translation vector. The initial actual point cloud is rotated and translated based on the rotation matrix and the three-dimensional translation vector to match the template point cloud, specifically: the actual point cloud is rotationally transformed based on the rotation matrix, and the actual point cloud is translationally transformed based on the three-dimensional translation amount.

[0059] In some embodiments, the preset initial transformation relationship is 0. The purpose of the present application is to optimize the initial transformation relationship using an objective function, to transform the actual point cloud converted by the optimized transformation relationship obtained by the previous iteration using the optimized transformation relationship obtained by the current iteration to match the template point cloud, and to make the matching accuracy between the actual point cloud converted according to the optimized transformation coefficients and the template point cloud higher, and the maximum distance deviation value between the actual point cloud and the template point cloud smaller.

[0060] Figure 2 A flowchart for constructing the objective function in another embodiment of the present application is shown in FIG. 6, which comprises the following steps: Figure 2 As shown, the template point cloud, the initial actual point cloud and the initial transformation relationship are used to construct the objective function, and the method further comprises the following steps:

[0061] In step 210, an initial rotation matrix between the template point cloud and the initial actual point cloud is determined based on the initial spatial rotation angle.

[0062] In step 220, a deviation equation is constructed based on the initial rotation matrix, the initial three-dimensional translation vector, the template point cloud and the initial actual point cloud, the deviation equation being used to represent the deviation between corresponding points in the template point cloud and the initial actual point cloud.

[0063] The deviation equation is expressed as:

[0064] Dev i = |R * p i + T - q i |

[0065] wherein R represents the initial rotation matrix, T represents the initial three-dimensional translation vector, p i represents the i-th point in the initial actual point cloud, and q i is the i-th point in the template point cloud.

[0066] In step 230, the objective function is constructed based on the deviation equation and the initial transformation relationship.

[0067] The objective function is used to define the problem of minimizing the maximum distance deviation value between the initial actual point cloud and the template point cloud when matched.

[0068] The objective function is expressed as:

[0069]

[0070] wherein x k represents the initial transformation relationship, including the initial spatial rotation angle (a, b, g) and the initial three-dimensional translation vector (T x , T y , T z ) in total six to be optimized parameters, and λ k , ξ k each represents a coefficient of the objective function, k represents the iteration number, and Φ i (x k , ξ k ) = Dev i - ξ k , I1 = {i e Φ i (x k , ξ k ) > 0, λki >0},I2={i∈Φ i (x k ,ξ k )>0,λ ki =0}.

[0071] In step 120, the objective function is iteratively optimized using a quasi-Newton algorithm to obtain an optimized transformation relationship. This optimized transformation relationship is used to transform the initial actual point cloud to obtain an optimized actual point cloud.

[0072] Figure 3 This illustration shows a flowchart of another embodiment of the present application, which uses a quasi-Newton algorithm to iteratively optimize the objective function and obtain the optimized transformation relationship. Figure 3 As shown, the objective function is iteratively optimized using a quasi-Newton algorithm to obtain the optimized transformation relationship, including the following steps:

[0073] In step 310, the coefficients of the objective function and the initial transformation relationship are initialized to obtain the initial values.

[0074] In some embodiments, during initialization (k=0), ξ0=0, λ0 is initialized to 1, and the initial transformation parameter x is initialized to 1. k The initial values ​​of the initial spatial rotation angle and the initial three-dimensional translation vector are both 0.

[0075] In step 320, the initial values ​​are substituted into the objective function for quasi-Newton iterative optimization, and the gradient of the objective function is calculated for each iteration;

[0076] Quasi-Newton methods are among the most efficient methods for solving nonlinear equations and performing optimization calculations. They are a class of Newton-type iterative methods that minimize computational complexity in each iteration while maintaining superlinear convergence. The algorithm calculates the gradient of the objective function at each iteration and uses the change in the gradient to determine if the iteration has converged.

[0077] In step 330, if the gradient of the objective function is less than the preset gradient value or the current iteration number is greater than the preset iteration number, then the optimized transformation relationship is output.

[0078] The initial transformation relationship is substituted into the objective function for iterative optimization. When the gradient of the objective function is less than the preset gradient value or the current iteration number is greater than the preset iteration number, the iteration stops and the optimized transformation relationship is output.

[0079] Figure 4 This illustration shows a flowchart of another embodiment of the present application for transforming an initial actual point cloud to obtain an optimized actual point cloud, as shown below. Figure 4 As shown, transforming the initial actual point cloud to obtain an optimized actual point cloud includes the following steps:

[0080] In step 410, the optimized transformation relationship is compared with the initial transformation relationship.

[0081] In step 420, if the difference between the optimized transformation relationship and the initial transformation relationship is not within the preset range, the initial actual point cloud is transformed using the optimized transformation relationship to obtain an optimized actual point cloud.

[0082] The transformation of the initial actual point cloud makes the maximum distance deviation between the initial actual point cloud and the template point cloud gradually smaller.

[0083] If the difference between the optimized transformation relationship and the initial transformation relationship is not within the preset range, it indicates that the iteration has not converged, and the quasi-Newton algorithm can optimize a more optimal transformation relationship.

[0084] In step 130, the target function is updated based on the optimized actual point cloud and the optimized transformation relationship to obtain a new target function, which is used as the target function for the next iteration optimization.

[0085] It should be noted that updating the target function includes updating the coefficients λ k , ξ k in the target function and updating the initial transformation relationship in the target function using the optimized transformation relationship.

[0086] The updated coefficients λ k , ξ k are specifically:

[0087]

[0088]

[0089]

[0090] The updated coefficients λ k , ξ k , transformation coefficients x k are brought into the target function to obtain a new target function. Before the quasi-Newton algorithm converges, the target function is updated using the optimized transformation relationship at each iteration, and then the new target function is used to continue multiple iteration optimizations, and the corresponding optimized transformation relationship is output.

[0091] The optimized transformation relationship output by each iteration optimization is used to transform the actual point cloud after the previous transformation, that is, the attitude of the actual point cloud is fine-tuned through the optimized transformation relationship, and the maximum distance deviation between the initial actual point cloud and the template point cloud is gradually reduced through each transformation.

[0092] Figure 5Fig. 4 shows a flowchart illustrating the steps of the method for performing quasi-Newton iterative optimization by substituting initial values into the objective function and calculating the gradient of the objective function at each iteration according to another embodiment of the present application, and the flowchart further comprises the following steps after the gradient of the objective function at each iteration is calculated: Figure 5

[0093] In step 510, if the gradient of the objective function is greater than or equal to the gradient threshold value or the current iteration number is less than the first preset iteration number, a search direction is calculated based on the gradient of the objective function at the current iteration and the initial matrix of the approximate Hessian matrix;

[0094] In step 520, a search step is calculated based on the search direction and the optimization transformation relationship of the current iteration;

[0095] In step 530, the approximate Hessian matrix is calculated based on the search direction, the search step, the optimization transformation relationship of the current iteration and the gradient of the objective function;

[0096] In step 540, the optimization transformation relationship of the current iteration is updated based on the approximate Hessian matrix;

[0097] In step 550, the gradient of the objective function of the next iteration is calculated based on the updated optimization transformation relationship until the gradient of the objective function is less than the gradient preset value or the current iteration number is greater than the first preset iteration number.

[0098] The essential idea of the quasi-Newton method is to use a positive definite matrix to approximate the inverse matrix of the Hessian matrix, to improve the defect that the inverse matrix of the complex Hessian matrix (i.e. the Hessian matrix) needs to be solved each time in the classical Newton method, to avoid the problem that the classical Newton iteration takes too much time to calculate the Hessian matrix, and to simplify the complexity of the operation.

[0099] In step 140, the iterative optimization is stopped when the objective function reaches the convergence criterion, and the target transformation relationship is determined based on the optimization transformation relationships obtained in the previous iterations, and the target transformation relationship is used to match the template point cloud and the initial actual point cloud, so that the maximum distance deviation value between the matched template point cloud and the initial actual point cloud is smaller.

[0100] It should be noted that the convergence criterion is that the difference between the optimization transformation relationship obtained by the iterative optimization and the initial transformation relationship is within a preset range, and in some embodiments, the preset initial transformation relationship is 0, and when the optimization transformation relationship is also 0, it indicates that the objective function cannot output a better value, at which time the objective function converges.

[0101] Figure 6 Fig. 5 shows a flowchart illustrating the steps of the method for matching the template point cloud and the initial actual point cloud using the target transformation relationship according to another embodiment of the present application, and the flowchart further comprises the following steps: Figure 6 ​As shown, the target transformation relationship is used to match the template point cloud and the initial actual point cloud, including the following steps:

[0102] In step 610, when the difference between the optimized transformation relationship and the initial transformation relationship is within the preset range, the iteration is stopped, and the optimized transformation relationship obtained through the iteration is output.

[0103] In step 620, the optimized transformation relationships obtained through the iteration are accumulated and summed to obtain the target transformation relationship.

[0104] In step 630, the spatial rotation angle in the target transformation relationship is converted into a rotation matrix.

[0105] In step 640, the initial actual point cloud is rotated and translated based on the rotation matrix and the three-dimensional translation in the target transformation relationship, so as to match the template point cloud.

[0106] In this application, the minimum problem of the maximum distance deviation between the actual point cloud and the template point cloud is expressed by constructing the objective function, the objective function is iteratively optimized by using the quasi-Newton algorithm, a plurality of optimized transformation relationships are output through multiple iterations, the target transformation relationship is determined based on the plurality of optimized transformation relationships, and the initial actual point cloud is converted based on the target transformation relationship, so that the matching accuracy between the converted actual point cloud and the template point cloud is higher and the maximum distance deviation is smaller.

[0107] Figure 7 The flowchart of the point cloud matching method based on quasi-Newton optimization of another embodiment of the application is shown as follows. Figure 7 As shown, the point cloud matching method based on quasi-Newton optimization includes the following steps:

[0108] In step 710, the template point cloud, the initial actual point cloud, and the initial transformation relationship between the template point cloud and the initial actual point cloud are obtained.

[0109] In step 720, the deviation equation is constructed based on the template point cloud, the initial actual point cloud, and the initial transformation relationship, and the objective function for minimizing the maximum distance deviation value available for quasi-Newton optimization is determined.

[0110] In step 730, the objective function is initialized, specifically, the initial transformation relationship and the coefficient in the objective function are initialized.

[0111] In step 740, the quasi-Newton iterative optimization process is performed according to the objective function, if the gradient of the objective function in the quasi-Newton iterative optimization process is less than a preset value or the current iteration number is greater than a preset iteration number, the optimized transformation relationship is output, otherwise the quasi-Newton method is used to iteratively update the optimized transformation relationship.

[0112] In step 750, the optimized transformation relationship is compared with the initial transformation relationship;

[0113] In step 760, when the difference between the optimized transformation relationship and the initial transformation relationship is not in the preset range, step 770 is performed; otherwise, step 780 is performed;

[0114] In step 770, the objective function is updated based on the optimized actual point cloud and the optimized transformation relationship, and after a new objective function is obtained, step 740 is continued to be performed;

[0115] In step 780, the optimized transformation relationships obtained in each iteration are accumulated and summed to obtain a target transformation relationship, the spatial rotation angle in the target transformation relationship is converted into a rotation matrix, and then the initial actual point cloud is rotated and translated based on the rotation matrix and the three-dimensional translation in the target transformation relationship, so as to match the template point cloud.

[0116] In the present application, the minimization problem of the maximum distance deviation between the actual point cloud and the template point cloud is expressed by constructing the objective function, the objective function is iteratively optimized by using the quasi-Newton algorithm, a plurality of optimized transformation relationships are output after multiple iterations, the target transformation relationship is determined based on the plurality of optimized transformation relationships, and the initial actual point cloud is converted based on the target transformation relationship, so that the matching accuracy between the converted actual point cloud and the template point cloud is higher and the maximum distance deviation is smaller.

[0117] In some embodiments, when the template point cloud and the actual point cloud are obtained, the normal vectors of each point cloud are also obtained, and then the deviation equation and the objective function are established based on the template point cloud with the normal vector and the initial actual point cloud, the quasi-Newton iterative optimization and the updating of the objective equation are performed, and the normal vector is introduced in the problem of minimizing the maximum distance deviation, so that the amount of calculation of the algorithm is reduced, and the optimal matching between the two point clouds with the normal vector is quickly and accurately realized.

[0118] Figure 8 The structure of the point cloud matching device based on quasi-Newton optimization provided in the present application is shown in the structure diagram, as shown in Figure 8 The point cloud matching device based on quasi-Newton optimization 800 includes an acquisition module 810, an iteration module 820, an objective function updating module 830, and a determination module 840. Wherein:

[0119] The acquisition module 810 is configured to acquire the template point cloud, the initial actual point cloud, and the initial transformation relationship between the template point cloud and the initial actual point cloud, and the template point cloud, the initial actual point cloud, and the initial transformation relationship are used to construct the objective function;

[0120] The iteration module 820 is configured to perform iterative optimization on the target function by using a quasi-Newton algorithm to obtain an optimized transformation relationship, and the optimized transformation relationship is used to transform the initial actual point cloud to obtain an optimized actual point cloud.

[0121] The target function updating module 830 is configured to update the target function based on the optimized actual point cloud and the optimized transformation relationship to obtain a new target function, and the new target function is used as the target function in the next iteration.

[0122] The determination module 840 is configured to stop the iteration when the target function reaches a convergence criterion, and determine a target transformation relationship based on the optimized transformation relationships obtained in the iterations, wherein the target transformation relationship is used to match the template point cloud and the initial actual point cloud, so that the maximum distance deviation value between the matched template point cloud and the initial actual point cloud is smaller.

[0123] In some embodiments, the transformation relationship includes a spatial rotation angle and a three-dimensional translation amount, the spatial rotation angle is used to determine a rotation matrix between the template point cloud and the actual point cloud, and the three-dimensional translation amount is used to represent a translation vector between the template point cloud and the actual point cloud.

[0124] In some embodiments, the obtaining module is further configured to determine an initial rotation matrix between the template point cloud and the initial actual point cloud based on the initial spatial rotation angle.

[0125] The deviation equation is used to represent the deviation between the corresponding points in the template point cloud and the initial actual point cloud.

[0126] The target function is constructed based on the deviation equation and the initial transformation relationship.

[0127] In some embodiments, the iteration module is further configured to initialize the coefficients of the target function and the initial transformation relationship to obtain initial values, substitute the initial values into the target function to perform quasi-Newton iterative optimization, calculate the gradient of the target function at each iteration, and output the optimized transformation relationship if the gradient of the target function is less than a gradient preset value or the current iteration number is greater than a first preset iteration number.

[0128] In some embodiments, the iteration module is further configured to transform the initial actual point cloud by using the optimized transformation relationship to obtain the optimized actual point cloud if the difference between the optimized transformation relationship and the initial transformation relationship is not within a preset range.

[0129] In some embodiments, the target function updating module is further configured to, if the target function gradient is greater than or equal to the gradient threshold or the current iteration number is less than the first preset iteration number, calculate a search direction based on the target function gradient at the current iteration and the initial matrix of the approximate Hessian matrix; calculate a search step based on the search direction and the optimization transformation relationship of the current iteration; calculate the approximate Hessian matrix based on the search direction, the search step, the optimization transformation relationship of the current iteration, and the target function gradient; and update the optimization transformation relationship of the current iteration based on the approximate Hessian matrix.

[0130] The target function gradient of the next iteration is calculated based on the updated optimization transformation relationship until the target function gradient is less than the gradient preset value or the current iteration number is greater than the preset iteration number.

[0131] In some embodiments, the convergence criterion is that a difference between the optimization transformation relationship obtained through iteration optimization and the initial transformation relationship is within a preset range.

[0132] In some embodiments, the determining module is further configured to: accumulate and sum the optimization transformation relationships obtained through all iterations to obtain a target transformation relationship.

[0133] In the present application, the minimization problem of the maximum distance deviation between the actual point cloud and the template point cloud is expressed by constructing a target function, the target function is iteratively optimized by using the quasi-Newton algorithm, a plurality of optimization transformation relationships are output through multiple iterations, the target transformation relationship is determined based on the plurality of optimization transformation relationships, and the initial actual point cloud is converted based on the target transformation relationship, so that the matching accuracy between the converted actual point cloud and the template point cloud is higher and the maximum distance deviation is smaller.

[0134] The above modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above modules.

[0135] In some embodiments, a terminal device is provided, including at least one processor and a memory, the memory is configured to store program instructions, and the processor is configured to invoke and execute the program instructions stored in the memory to enable the terminal device to perform the point cloud matching method based on quasi-Newton optimization described above. The implementation principle and technical effects are similar to those of the above method embodiments, and will not be repeated here.

[0136] In some embodiments, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium stores instructions, which, when executed on a computer, cause the computer to perform the point cloud matching method based on quasi-Newton optimization described above. The implementation principle and technical effects are similar to those of the above method embodiments, and will not be repeated here.

[0137] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0138] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. Any reference to memory, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided by the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided by the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0139] The technical features of the above embodiments can be combined in any way. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0140] The above embodiments only express several implementation ways of the present application, and the description is specific and detailed, but it should not be understood as a limitation to the patent scope of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A point cloud matching method based on quasi-Newtonian optimization, characterized in that, The method includes: Obtain a template point cloud, an initial actual point cloud, and an initial transformation relationship between the template point cloud and the initial actual point cloud. The template point cloud, the initial actual point cloud, and the initial transformation relationship are used to construct the objective function. The objective function is iteratively optimized using a quasi-Newton algorithm to obtain an optimized transformation relationship, which is then used to transform the initial actual point cloud to obtain an optimized actual point cloud. The objective function is updated based on the optimized actual point cloud and the optimized transformation relationship to obtain a new objective function, which is used as the objective function for the next iteration. The iteration stops when the objective function reaches the convergence criterion, and the target transformation relationship is determined based on the optimized transformation relationship obtained in each iteration. The target transformation relationship is used to match the template point cloud and the initial actual point cloud so that the maximum distance deviation between the template point cloud and the initial actual point cloud is smaller after matching. The step of iteratively optimizing the objective function using a quasi-Newton algorithm to obtain the optimized transformation relationship further includes: Initialize the coefficients of the objective function and the initial transformation relationship to obtain initial values; Substitute the initial value into the objective function to perform quasi-Newton iterative optimization, and calculate the gradient of the objective function at each iteration; If the gradient of the objective function is less than the preset gradient value or the current iteration number is greater than the first preset iteration number, then the optimized transformation relationship is output; After substituting the initial value into the objective function for quasi-Newton iterative optimization and calculating the gradient of the objective function at each iteration, the process further includes: If the gradient of the objective function is greater than or equal to the gradient threshold or the current iteration number is less than the first preset iteration number, then the search direction is calculated based on the gradient of the objective function at the current iteration and the initial matrix of the approximate Hessian matrix; The search step size is calculated based on the search direction and the optimization transformation relationship of the current iteration; The approximate Hessian matrix is ​​calculated based on the search direction, the search step size, the optimization transformation relationship of the current iteration, and the gradient of the objective function; The optimization transformation relationship of the current iteration is updated based on the approximate Hessian matrix; Calculate the gradient of the objective function for the next iteration based on the updated optimization transformation relationship, until the gradient of the objective function is less than the preset gradient value or the current iteration number is greater than the preset iteration number; Determining the target transformation relationship based on the optimized transformation relationship obtained from each iteration further includes: The optimized transformation relationships obtained from each iteration are summed to obtain the target transformation relationship.

2. The point cloud matching method based on quasi-Newton optimization as described in claim 1, characterized in that, The transformation relationship includes a spatial rotation angle and a three-dimensional translation amount. The spatial rotation angle is used to determine the rotation matrix between the template point cloud and the actual point cloud, and the three-dimensional translation amount is used to characterize the translation vector between the template point cloud and the actual point cloud.

3. The point cloud matching method based on quasi-Newton optimization as described in claim 1, characterized in that, The initial transformation relationship includes an initial spatial rotation angle and an initial three-dimensional translation vector. The template point cloud, the initial actual point cloud, and the initial transformation relationship are used to construct the objective function, and further include: Determine the initial rotation matrix between the template point cloud and the initial actual point cloud based on the initial spatial rotation angle; Based on the initial rotation matrix, the initial 3D translation vector, the template point cloud, and the initial actual point cloud, a deviation equation is constructed. The deviation equation is used to characterize the deviation between corresponding points in the template point cloud and the initial actual point cloud. The objective function is constructed based on the deviation equation and the initial transformation relationship.

4. The point cloud matching method based on quasi-Newton optimization as described in claim 1, characterized in that, The optimized transformation relationship is used to transform the initial actual point cloud to obtain an optimized actual point cloud, and further includes: If the difference between the optimized transformation relationship and the initial transformation relationship is not within a preset range, the initial actual point cloud is transformed using the optimized transformation relationship to obtain the optimized actual point cloud.

5. The point cloud matching method based on quasi-Newton optimization as described in claim 1, characterized in that, The convergence criterion is that the difference between the optimized transformation relationship obtained through iterative optimization and the initial transformation relationship is within a preset range.

6. A point cloud matching device based on quasi-Newtonian optimization, characterized in that, include: The acquisition module is used to acquire a template point cloud, an initial actual point cloud, and an initial transformation relationship between the template point cloud and the initial actual point cloud. The template point cloud, the initial actual point cloud, and the initial transformation relationship are used to construct the objective function. An iterative module is used to iteratively optimize the objective function using a quasi-Newton algorithm to obtain an optimized transformation relationship, which is used to transform the initial actual point cloud to obtain an optimized actual point cloud. The objective function update module is used to update the objective function based on the optimized actual point cloud and the optimized transformation relationship to obtain a new objective function, which is used as the objective function for the next iteration. The determination module is used to stop iterating when the objective function reaches the convergence criterion, and to determine the target transformation relationship based on the optimized transformation relationship obtained in each iteration, wherein the target transformation relationship is used to match the template point cloud and the initial actual point cloud so that the maximum distance deviation between the template point cloud and the initial actual point cloud is smaller after matching; The step of iteratively optimizing the objective function using a quasi-Newton algorithm to obtain the optimized transformation relationship further includes: Initialize the coefficients of the objective function and the initial transformation relationship to obtain initial values; Substitute the initial value into the objective function to perform quasi-Newton iterative optimization, and calculate the gradient of the objective function at each iteration; If the gradient of the objective function is less than the preset gradient value or the current iteration number is greater than the first preset iteration number, then the optimized transformation relationship is output; After substituting the initial value into the objective function for quasi-Newton iterative optimization and calculating the gradient of the objective function at each iteration, the process further includes: If the gradient of the objective function is greater than or equal to the gradient threshold or the current iteration number is less than the first preset iteration number, then the search direction is calculated based on the gradient of the objective function at the current iteration and the initial matrix of the approximate Hessian matrix; The search step size is calculated based on the search direction and the optimization transformation relationship of the current iteration; The approximate Hessian matrix is ​​calculated based on the search direction, the search step size, the optimization transformation relationship of the current iteration, and the gradient of the objective function; The optimization transformation relationship of the current iteration is updated based on the approximate Hessian matrix; Calculate the gradient of the objective function for the next iteration based on the updated optimization transformation relationship, until the gradient of the objective function is less than the preset gradient value or the current iteration number is greater than the preset iteration number; Determining the target transformation relationship based on the optimized transformation relationship obtained from each iteration further includes: The optimized transformation relationships obtained from each iteration are summed to obtain the target transformation relationship.

7. The point cloud matching device based on quasi-Newton optimization as described in claim 6, characterized in that, The transformation relationship includes a spatial rotation angle and a three-dimensional translation amount. The spatial rotation angle is used to determine the rotation matrix between the template point cloud and the actual point cloud, and the three-dimensional translation amount is used to characterize the translation vector between the template point cloud and the actual point cloud.

Images