A method, system and terminal device for point cloud registration using random coefficients

Abstract

The application discloses a method, a system and a terminal device for point cloud registration using random coefficients, and the method comprises the following steps: acquiring a point cloud data set, and performing data preprocessing on the point cloud data set to obtain a point cloud pair to be registered; using a random coefficient method to perform point cloud filling on the point cloud pair to be registered to obtain a test point cloud data set; inputting the test point cloud data set into a point cloud registration model to perform point cloud registration, and obtaining a point cloud registration coincidence degree. In the application, the idea of random points is used, points selected randomly from a point cloud with a small number of points are used to add some points to the point cloud with a small number of points, so that the number of points in the point cloud pair to be registered is consistent, and then the point cloud registration is performed; the method of random points ensures that the newly added points do not participate in registration as corresponding points, and simultaneously solves the problem that the number of points in two point clouds is not equal; after obtaining the point cloud pair with equal number of points, fast and accurate registration can be performed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of point cloud registration, and in particular to a method, system and terminal device for point cloud registration using random coefficients. BACKGROUND

[0002] 3D point cloud registration technology refers to a process of aligning two point cloud information of the same object or environment acquired at different positions, different angles and different sensors by coordinate transformation, and has been widely applied in automatic driving, 3D reconstruction, robot positioning and many other fields. Existing point cloud registration methods are mainly divided into traditional methods and deep learning-based methods. The traditional method finds the corresponding relationship between points of the point cloud to be registered and the target point cloud by manually making point cloud features, while the deep learning algorithm learns the local or global features of the point cloud through MLP, GNN, transformer and other networks, and then obtains the transformation relationship between the two point clouds. However, in real-world scenarios, due to different sensors, the acquired point cloud density is different, and the point number is also not equal. Even if it is the same sensor, due to different scanning times and environments, the point number acquired is also not equal. However, many deep learning models require the point number of the point cloud to be registered and the target point cloud to be consistent. Therefore, the traditional method is time-consuming and inefficient, and the deep learning-based method has great limitations in use due to the point number consistency requirement.

[0003] There are two cases for the point number inconsistency between the point cloud to be registered and the target point cloud. One is that the point number of the point cloud to be registered is more than that of the target point cloud, and the other is that the point number of the point cloud to be registered is less than that of the target point cloud. Whether the point number of the point cloud to be registered is changed or the point number of the target point cloud is changed, the purpose is to keep the point number of the point cloud to be registered and the target point cloud consistent. The common processing method is downsampling and upsampling. Downsampling is to randomly remove part of the points with more point numbers to keep the point numbers of the two point clouds consistent. However, this processing is fatal to "partial point cloud registration". The partial point cloud registration problem is caused by the fact that the point cloud is only partially overlapped, and the point number of the overlapped part is not large. If a part of the points is removed, the point number of the overlapped part will decrease, which may not be conducive to the point cloud registration. Upsampling is to increase a part of points of the point cloud with less point number by using a certain method. The current method includes "linear interpolation" and "quadratic interpolation". This method can convert sparse point cloud into dense point cloud and increase the point number arbitrarily. However, in the registration process, the added points may become "corresponding points" between the two point clouds. However, the added points are not the existing points, which may result in an incorrect registration.

[0004] Therefore, the prior art still needs to be improved. SUMMARY

[0005] The technical problem solved by the present application is to provide a method, system and terminal device for point cloud registration using random coefficients to solve the problem of unequal point cloud points in the point cloud registration process in the prior art.

[0006] The technical solution adopted by the present application to solve the problem is as follows:

[0007] In a first aspect, the present application provides a method for point cloud registration using random coefficients, wherein the method comprises:

[0008] Obtaining a point cloud data set and performing data preprocessing on the point cloud data set to obtain a pair of point clouds to be registered;

[0009] Performing point cloud filling on the pair of point clouds to be registered using a random coefficient method to obtain a test point cloud data set;

[0010] Inputting the test point cloud data set into a point cloud registration model for point cloud registration to obtain a point cloud registration coincidence degree.

[0011] In an implementation manner, the data preprocessing on the point cloud data set to obtain a pair of point clouds to be registered specifically comprises:

[0012] Performing primary cutting on the point cloud data set using a first cutting parameter to obtain a first target registration point cloud;

[0013] Performing secondary cutting on the point cloud data set using a second cutting parameter to obtain a first point cloud to be registered;

[0014] Performing third cutting on the first target registration point cloud and the first point cloud to be registered using a third cutting parameter to obtain a pair of point clouds to be registered, wherein the pair of point clouds to be registered comprises a target registration point cloud and a point cloud to be registered.

[0015] In an implementation manner, the point cloud filling on the pair of point clouds to be registered using a random coefficient method to obtain a test point cloud data set specifically comprises:

[0016] Comparing the number of point cloud points in the target registration point cloud with the number of point cloud points in the point cloud to be registered;

[0017] If the number of point cloud points in the target registration point cloud is greater than the number of point cloud points in the point cloud to be registered, performing point cloud filling on the point cloud to be registered so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered;

[0018] If the number of point cloud points in the target registration point cloud is less than the number of point cloud points in the point cloud to be registered, performing point cloud filling on the target registration point cloud so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered.

[0019] In an implementation manner, the point cloud filling on the to-be-registered point cloud pair specifically comprises:

[0020] calculating a difference value between a point cloud point number in the target registration point cloud and a point cloud point number in the to-be-registered point cloud;

[0021] randomly selecting, in the to-be-registered point cloud, a to-be-registered point cloud with the same point number as the difference value as a first filling point cloud;

[0022] adding the first filling point cloud to the to-be-registered point cloud.

[0023] In an implementation manner, the point cloud filling on the target registration point cloud specifically comprises:

[0024] obtaining a difference value between a point cloud point number in the target registration point cloud and a point cloud point number in the to-be-registered point cloud;

[0025] randomly selecting, in the target registration point cloud, a target registration point cloud with the same point number as the difference value as a second filling point cloud;

[0026] adding the second filling point cloud to the target registration point cloud.

[0027] In an implementation manner, the first clipping parameter is the same as the second clipping parameter.

[0028] In an implementation manner, the third clipping parameter is a random number between 0.65 and 0.75.

[0029] In a second aspect, an embodiment of the present application further provides a system for point cloud registration using random coefficients, comprising:

[0030] a data acquisition module, configured to acquire a point cloud dataset, and perform data preprocessing on the point cloud dataset to obtain a to-be-registered point cloud pair;

[0031] a data processing module, configured to perform point cloud filling on the to-be-registered point cloud pair using a random coefficient method to obtain a test point cloud dataset;

[0032] a point cloud registration module, configured to input the test point cloud dataset into a point cloud registration model to perform point cloud registration, and obtain a point cloud registration coincidence degree.

[0033] In an implementation manner, the data acquisition module comprises:

[0034] a first clipping unit, configured to perform primary clipping on the point cloud dataset using a first clipping parameter to obtain a first target registration point cloud;

[0035] a second clipping unit, configured to perform second clipping on the point cloud dataset by using a second clipping parameter, to obtain a first to-be-registered point cloud;

[0036] a third clipping unit, configured to perform third clipping on the first target-registered point cloud and the first to-be-registered point cloud by using a third clipping parameter, to obtain a to-be-registered point cloud pair, the to-be-registered point cloud pair including a target-registered point cloud and a to-be-registered point cloud.

[0037] In an implementation manner, the data processing module includes:

[0038] a comparison unit, configured to compare a point cloud point number in the target-registered point cloud with a point cloud point number in the to-be-registered point cloud;

[0039] a first point cloud filling unit, configured to perform point cloud filling on the to-be-registered point cloud, if the point cloud point number in the target-registered point cloud is greater than the point cloud point number in the to-be-registered point cloud, so that the point cloud point number in the target-registered point cloud is the same as the point cloud point number in the to-be-registered point cloud;

[0040] a second point cloud filling unit, configured to perform point cloud filling on the target-registered point cloud, if the point cloud point number in the target-registered point cloud is less than the point cloud point number in the to-be-registered point cloud, so that the point cloud point number in the target-registered point cloud is the same as the point cloud point number in the to-be-registered point cloud.

[0041] In an implementation manner, the first point cloud filling unit includes:

[0042] a first difference calculation sub-unit, configured to calculate a difference between the point cloud point number in the target-registered point cloud and the point cloud point number in the to-be-registered point cloud;

[0043] a first random point selection sub-unit, configured to randomly select, in the to-be-registered point cloud, a to-be-registered point cloud with the same point number as the difference as a first filling point cloud;

[0044] a first random point filling sub-unit, configured to add the first filling point cloud to the to-be-registered point cloud.

[0045] In an implementation manner, the second point cloud filling unit includes:

[0046] a difference acquisition sub-unit, configured to acquire a difference between the point cloud point number in the target-registered point cloud and the point cloud point number in the to-be-registered point cloud;

[0047] a second random point selection sub-unit, configured to randomly select, in the target-registered point cloud, a target-registered point cloud with the same point number as the difference as a second filling point cloud;

[0048] The second random point filling subunit is configured to fill the second added point cloud into the target registration point cloud.

[0049] In a third aspect, the present application also provides a terminal device, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by one or more processors, and the one or more programs executed by the one or more processors comprise the method for point cloud registration using random coefficients according to any one of the above.

[0050] In a fourth aspect, the present application also provides a non-transitory computer readable storage medium, when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the method for point cloud registration using random coefficients according to any one of the above.

[0051] The present application provides a method, system and terminal device for point cloud registration using random coefficients. The method first acquires a point cloud data set, and performs data preprocessing on the point cloud data set to obtain a pair of point clouds to be registered. Then, the random coefficient method is used to fill the point clouds to be registered to obtain a test point cloud data set. Finally, the test point cloud data set is input into a point cloud registration model for point cloud registration to obtain a point cloud registration coincidence degree. In the present application, the idea of random points is used to add a part of points to the point cloud with fewer points by randomly selecting points, so that the number of point cloud points in the pair of point clouds to be registered is consistent, and then the point cloud registration is performed. The method of random points ensures that the newly added points do not participate in registration as corresponding points, and also solves the problem of unequal number of points in two point clouds. After obtaining the point clouds with equal number of points, fast and accurate registration can be performed. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0053] Figure 1 is a flowchart of the method for point cloud registration based on random points provided by the embodiments of the present application.

[0054] Figure 2 is a system principle block diagram of the method for point cloud registration based on random points provided by the embodiments of the present application.

[0055] Figure 3 is an internal structure principle block diagram of the terminal device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0056] The present application will be further described in details below with reference to the drawings and examples. It should be understood that the specific examples described herein are intended to explain the present application and are not intended to limit the present application.

[0057] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement condition, etc. between components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional indications will also change accordingly.

[0058] In the prior art, the point cloud registration is basically performed by upsampling or downsampling to make the number of the point cloud to be registered and the target point cloud consistent. However, this will cause the overlapping part of the point cloud to be reduced or the non-existing point cloud to be increased to participate in the registration, so that the registration process is difficult to perform or an incorrect registration is generated.

[0059] To solve the problems in the prior art, the embodiment provides a method for point cloud registration based on random points, by which the point cloud registration can be realized. The method not only has high registration efficiency, but also has high registration quality. In specific implementation, first, a point cloud dataset is acquired, and data preprocessing is performed on the point cloud dataset to obtain a point cloud pair to be registered. Then, a random coefficient method is used to fill the point cloud of the point cloud pair to be registered, so that the number of point cloud points in the target point cloud and the number of point cloud points in the point cloud to be registered are consistent, and a test point cloud dataset is obtained. Finally, the test point cloud dataset is input into a point cloud registration model to perform point cloud registration, and a point cloud registration coincidence degree is obtained. By randomly adding point clouds to make the number of point cloud points consistent, the point cloud registration is performed, which can improve the effect of point cloud registration and also improve the speed of point cloud registration.

[0060] For example, the radar obtains a set of point cloud data, and the system first performs data preprocessing on the point cloud data set to obtain a point cloud pair to be registered. At this time, the number of point cloud points in the point cloud pair to be registered is not equal, and it is necessary to process it to make the number of point cloud points the same to further register the point cloud. By adding a part of random points to the point cloud with fewer point cloud points, the number of point cloud points in the point cloud pair to be registered is the same. For example, if the number of point cloud points in the target registration point cloud is less than the number of point cloud points in the point cloud to be registered, the target registration point cloud needs to be increased in point cloud to make it the same as the number of point cloud points in the point cloud to be registered. Finally, after obtaining the test point cloud data set with the same number of point cloud points, it is input into the point cloud registration model to automatically register the point cloud and obtain the coincidence degree between the point clouds. By increasing the random points, the number of point cloud points in the point cloud pair to be registered is kept consistent, which can effectively solve the problem of inconsistent number of points, improve the effect of point cloud registration, and also improve the speed of point cloud registration.

[0061] Exemplary method

[0062] The embodiment provides a method for point cloud registration based on random points, which can be applied to a terminal device. Specifically, as shown in the figure, Figure 1 The method comprises the following steps:

[0063] In step S100, a point cloud data set is obtained, and data preprocessing is performed on the point cloud data set to obtain a point cloud pair to be registered.

[0064] In this embodiment, the point cloud data set is taken from the ModelNet40 data set, which is an open source data set. By preprocessing the collected point cloud data, a point cloud pair to be registered with different numbers of points can be obtained, which facilitates random point cloud increase processing to make the number of point cloud points in the point cloud pair to be registered equal, and finally performs fast point cloud registration through a point cloud registration model.

[0065] In one implementation mode, a first clipping parameter is used to perform initial clipping on the point cloud data set to obtain a first target registration point cloud; a second clipping parameter is used to perform secondary clipping on the point cloud data set to obtain a first point cloud to be registered; and a third clipping parameter is used to perform third clipping on the first target registration point cloud and the first point cloud to be registered to obtain a point cloud pair to be registered, which includes a target registration point cloud and a point cloud to be registered. After preprocessing the point cloud data set, a point cloud pair to be registered with different numbers of points can be obtained.

[0066] In a specific implementation, the first cutting parameter is used to perform a first cutting on the collected point cloud dataset to obtain a first target registration point cloud, for example, 30% of the points are cut off and only 70% of the points are retained as the first target registration point cloud, and the remaining point cloud may only be a part of the original point cloud. Then, a second cutting parameter is used to perform a second cutting on the collected point cloud dataset to obtain a first registration point cloud, and the second cutting also cuts off 30% of the points and only retains 70% of the points as the first registration point cloud. For example, the first cutting of 30% of the points cuts off the legs, and the second cutting of 30% of the points may cut off the arms. The first cutting and the second cutting are independent and do not affect each other, so the cut-off parts are different, and the retained parts are also different. After obtaining the first target point cloud and the first registration point cloud, a third cutting is performed on them by selecting a random number, and a pair of registration point clouds with different point numbers can be obtained. The first cutting parameter and the second cutting parameter are the same. The third cutting parameter is a random number between 0.65 and 0.75. After multiple cuttings of the point cloud dataset, a pair of registration point clouds with different point numbers is obtained.

[0067] In a specific implementation, the first cutting and the second cutting can obtain a first target registration point cloud and a first registration point cloud, respectively, wherein the first cutting parameter and the second cutting parameter are the same, for example, both are 0.7, that is, 70% of the point cloud is cut off. The third cutting is a random value between 0.65 and 0.75. In experiments, the average of the third cutting parameter is 0.7, and the purpose is to cut off different numbers of point clouds each time, in order to create a more realistic scenario, because the number of point clouds obtained by each scan of the real radar is different.

[0068] In an implementation, after obtaining the first target registration point cloud, rotation, translation transformation and other operations are performed on it, and three rotation matrices are randomly generated, which correspond to X, Y and Z axes, for example, the rotation matrix of the X axis is to rotate the point cloud around the X axis, and the Y axis and the Z axis are the same. The translation vector is a three-dimensional vector, which also corresponds to the X, Y and Z axes. After the rotation and translation transformation of the first target registration point cloud are performed, it is applied to the first registration point cloud, and a registration point cloud pair for training is obtained.

[0069] In step S200, a random coefficient method is used to fill the registration point cloud pair to obtain a test point cloud dataset.

[0070] In the embodiment, after obtaining the pair of point clouds to be registered, the point cloud increasing processing is performed to obtain a test point cloud data set with the same number of points. When the number of points is the same, the point cloud registration is performed. The random points are randomly selected from the point cloud with a smaller number of points. For example, if the number of points of the point cloud to be registered is smaller, a part of points is randomly selected from the point cloud to be registered and added to the point cloud to be registered. The newly added points and the original points in the point cloud to be registered have the same spatial coordinates, and the spatial distribution of the point cloud is not changed. At the same time, the points that do not exist in the point cloud are not generated. For the case that the number of points of the target registration point cloud is smaller, the same processing idea is adopted. The purpose is to make the number of points of the point cloud to be registered and the number of points of the target registration point cloud equal.

[0071] In the specific implementation, firstly, the number of points of the target registration point cloud is compared with the number of points of the point cloud to be registered. If the number of points of the target registration point cloud is greater than the number of points of the point cloud to be registered, the point cloud to be registered is filled to make the number of points of the target registration point cloud and the number of points of the point cloud to be registered the same. If the number of points of the target registration point cloud is smaller than the number of points of the point cloud to be registered, the target registration point cloud is filled to make the number of points of the target registration point cloud and the number of points of the point cloud to be registered the same. Through the increasing of the number of points of the point cloud with a smaller number of points, the number of points of the point cloud to be registered and the number of points of the target registration point cloud are made the same, and the point cloud registration is realized.

[0072] In the specific implementation, the difference between the number of points of the target registration point cloud and the number of points of the point cloud to be registered is calculated. The point cloud to be registered with the same number of points as the difference is randomly selected from the point cloud to be registered as a first filling point cloud. The first filling point cloud is added to the point cloud to be registered as the point cloud to be registered in the test point cloud data set.

[0073] In the specific implementation, the difference between the number of points of the target registration point cloud and the number of points of the point cloud to be registered is obtained. The target registration point cloud with the same number of points as the difference is randomly selected from the target registration point cloud as a second filling point cloud. The second filling point cloud is added to the target registration point cloud as the target registration point cloud in the test point cloud data set.

[0074] The pair of point clouds to be registered after the random point increasing processing has solved the problem that the number of points is not equal in the point cloud registration, and the point cloud registration model can be further used for the point cloud registration.

[0075] In step S300, the test point cloud data set is input into the point cloud registration model for the point cloud registration, and the point cloud registration coincidence degree is obtained.

[0076] In the embodiment, after the test point cloud dataset is processed, the point cloud registration model is input into the point cloud registration model for point cloud registration, and the point cloud registration coincidence degree is obtained.

[0077] In the application, the increased random points do not affect the registration. In the experiment, the case where the clipping parameter is fixed as 0.7 is compared. The experimental results show that the influence of the random points on the registration can be almost ignored. In the case where the parameter is fixed as 0.7, the experimental results are as follows: in the registration of the partial point cloud with noise, for the registration of the category that does not appear in the training set, the angle error is 0.79 degrees, and the translation error is 0.0090 degrees; for the category that appears in the training set, the angle error reaches 0.38 degrees, and the translation error reaches 0.0041 degrees. In the case where the clipping parameter between (0.65, 0.75) in the application is adopted, the experimental results are as follows: in the registration of the partial point cloud with noise, for the registration of the category that does not appear in the training set, the angle error is 0.80 degrees, and the translation error is 0.0091 degrees; for the category that appears in the training set, the angle error reaches 0.39 degrees, and the translation error reaches 0.0042 degrees. The experimental results show that the random points in the application almost do not affect the registration effect, and can effectively solve the problem of unequal point cloud points.

[0078] Exemplary system

[0079] As shown in Figure 2 The embodiment of the application also provides a system for point cloud registration using random coefficients, which comprises a data acquisition module 10, a data processing module 20 and a point cloud registration module 30. Specifically, the data acquisition module 10 is used for acquiring a point cloud dataset and performing data preprocessing on the point cloud dataset to obtain a point cloud pair to be registered; the data processing module 20 is used for performing point cloud filling on the point cloud pair to be registered by using a random coefficient method to obtain a test point cloud dataset; and the point cloud registration module 30 is used for inputting the test point cloud dataset into a point cloud registration model for point cloud registration to obtain a point cloud registration coincidence degree.

[0080] In an implementation manner, the data acquisition module comprises:

[0081] A first clipping unit is configured to perform primary clipping on the point cloud dataset by using a first clipping parameter to obtain a first target registration point cloud;

[0082] A second clipping unit is configured to perform secondary clipping on the point cloud dataset by using a second clipping parameter to obtain a first point cloud to be registered;

[0083] A third clipping unit is configured to perform third clipping on the first target registration point cloud and the first point cloud to be registered by using a third clipping parameter to obtain a point cloud pair to be registered, wherein the point cloud pair to be registered comprises a target registration point cloud and a point cloud to be registered.

[0084] In an implementation manner, the data processing module comprises:

[0085] a comparison unit configured to compare the number of point cloud points in the target registration point cloud with the number of point cloud points in the point cloud to be registered;

[0086] a first point cloud filling unit configured to, if the number of point cloud points in the target registration point cloud is greater than the number of point cloud points in the point cloud to be registered, fill the point cloud to be registered so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered;

[0087] a second point cloud filling unit configured to, if the number of point cloud points in the target registration point cloud is less than the number of point cloud points in the point cloud to be registered, fill the target registration point cloud so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered.

[0088] In an implementation manner, the first point cloud filling unit comprises:

[0089] a first difference calculation sub-unit configured to calculate the difference between the number of point cloud points in the target registration point cloud and the number of point cloud points in the point cloud to be registered;

[0090] a first random point selection sub-unit configured to randomly select, in the point cloud to be registered, a number of point cloud points to be registered equal to the difference as a first filling point cloud;

[0091] a first random point filling sub-unit configured to add the first filling point cloud to the point cloud to be registered.

[0092] In an implementation manner, the second point cloud filling unit comprises:

[0093] a difference acquisition sub-unit configured to acquire the difference between the number of point cloud points in the target registration point cloud and the number of point cloud points in the point cloud to be registered;

[0094] a second random point selection sub-unit configured to randomly select, in the target registration point cloud, a number of target registration point clouds equal to the difference as a second filling point cloud;

[0095] a second random point filling sub-unit configured to add the second filling point cloud to the target registration point cloud.

[0096] Based on the above embodiments, the application further provides a terminal device, and a principle block diagram of the terminal device can be shown in FIG. 3. The terminal device can comprise one or more processors 100( Figure 3The memory 101 and the computer program 102 stored in the memory 101 and executable on the one or more processors 100, for example, a program of the method for point cloud registration using random coefficients. The one or more processors 100 can implement each step in the embodiment of the method for point cloud registration using random coefficients when executing the computer program 102. Alternatively, the one or more processors 100 can implement the functions of each module / unit in the embodiment of the system for point cloud registration using random coefficients when executing the computer program 102, which is not limited here.

[0097] In one embodiment, the processor 100 can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0098] In one embodiment, the memory 101 can be an internal storage unit of the electronic device, such as a hard disk or a memory of the electronic device. The memory 101 can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 101 can include both the internal storage unit and the external storage device of the electronic device. The memory 101 is used to store computer programs and other programs and data required by the terminal device. The memory 101 can also be used to temporarily store data that has been output or will be output.

[0099] Those skilled in the art can understand that, Figure 3 The principle block diagram shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the terminal device to which the scheme of the present application is applied. The specific terminal device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0100] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, operating database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0101] In summary, the present application discloses a method, system and terminal device for point cloud registration using random coefficients, the method comprising: obtaining a point cloud data set, and pre-processing the point cloud data set to obtain a pair of point clouds to be registered; using a random coefficient method to fill the point clouds to be registered to obtain a test point cloud data set; inputting the test point cloud data set into a point cloud registration model for point cloud registration to obtain a point cloud registration coincidence degree. In the present application, the idea of random points is used to add a part of points to the point cloud with fewer points by randomly selecting points, so that the number of point cloud points in the pair of point clouds to be registered is consistent, and then the point cloud registration is performed. The method of random points ensures that the newly added points do not participate in registration as corresponding points, and also solves the problem of unequal number of points in two point clouds. After obtaining the point cloud with equal number of points, fast and accurate registration can be performed.

[0102] It should be understood that the application of the present application is not limited to the above examples, and those skilled in the art can make improvements or changes according to the above description, and all these improvements and changes shall belong to the protection scope of the claims of the present application.

Claims

1. A method for point cloud registration using random coefficients, characterized in that, The method includes: Acquire a point cloud dataset and perform data preprocessing on the point cloud dataset to obtain point cloud pairs to be registered; The point cloud pairs to be registered are filled with point clouds using a random coefficient method to obtain a test point cloud dataset. The test point cloud dataset is input into the point cloud registration model for point cloud registration to obtain the point cloud registration overlap. The step of preprocessing the point cloud dataset to obtain point cloud pairs to be registered specifically includes: The point cloud dataset is initially cropped using the first cropping parameter to obtain the first target registration point cloud. The point cloud dataset is cropped a second time using the second cropping parameter to obtain the first point cloud to be registered. The first target registration point cloud and the first registration point cloud are cropped three times using a third cropping parameter to obtain a registration point cloud pair, which includes the target registration point cloud and the registration point cloud. The step of filling the point cloud pairs to be registered using a random coefficient method to obtain a test point cloud dataset specifically includes: Compare the number of point clouds in the target registration point cloud with the number of point clouds in the point cloud to be registered; If the number of point cloud points in the target registration point cloud is greater than the number of point cloud points in the point cloud to be registered, then the point cloud to be registered is filled with point cloud points so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered. If the number of point cloud points in the target registration point cloud is less than the number of point cloud points in the point cloud to be registered, then the target registration point cloud is filled with point cloud points so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered. The point cloud filling process for the point cloud pair to be registered specifically includes: Calculate the difference between the number of point cloud points in the target registration point cloud and the number of point cloud points in the point cloud to be registered; Randomly select a point cloud with the same number of points as the difference from the point cloud to be registered as the first filled point cloud; Add the first filled point cloud to the point cloud to be registered; The point cloud filling process for the target registration point cloud specifically includes: Obtain the difference between the number of point clouds in the target registration point cloud and the number of point clouds in the point cloud to be registered; Randomly select a target registration point cloud with the same number of points as the difference from the target registration point cloud as the second filling point cloud; The second filled point cloud is added to the target registration point cloud.

2. The method for point cloud registration using random coefficients according to claim 1, characterized in that, The first cutting parameter is the same as the second cutting parameter.

3. The method for point cloud registration using random coefficients according to claim 1, characterized in that, The third trimming parameter is a random number between 0.65 and 0.

75.

4. A system for point cloud registration using random coefficients, characterized in that, The system includes: The data acquisition module is used to acquire a point cloud dataset and perform data preprocessing on the point cloud dataset to obtain point cloud pairs to be registered. The data processing module is used to fill the point cloud pairs to be registered using a random coefficient method to obtain a test point cloud dataset. The point cloud registration module is used to input the test point cloud dataset into the point cloud registration model for point cloud registration and obtain the point cloud registration overlap. The step of preprocessing the point cloud dataset to obtain point cloud pairs to be registered specifically includes: The point cloud dataset is initially cropped using the first cropping parameter to obtain the first target registration point cloud. The point cloud dataset is cropped a second time using the second cropping parameter to obtain the first point cloud to be registered. The first target registration point cloud and the first registration point cloud are cropped three times using a third cropping parameter to obtain a registration point cloud pair, which includes the target registration point cloud and the registration point cloud. The step of filling the point cloud pairs to be registered using a random coefficient method to obtain a test point cloud dataset specifically includes: Compare the number of point clouds in the target registration point cloud with the number of point clouds in the point cloud to be registered; If the number of point cloud points in the target registration point cloud is greater than the number of point cloud points in the point cloud to be registered, then the point cloud to be registered is filled with point cloud points so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered. If the number of point cloud points in the target registration point cloud is less than the number of point cloud points in the point cloud to be registered, then the target registration point cloud is filled with point cloud points so that the number of point cloud points in the target registration point cloud is the same as the number of point cloud points in the point cloud to be registered. The point cloud filling process for the point cloud pair to be registered specifically includes: Calculate the difference between the number of point cloud points in the target registration point cloud and the number of point cloud points in the point cloud to be registered; Randomly select a point cloud with the same number of points as the difference from the point cloud to be registered as the first filled point cloud; Add the first filled point cloud to the point cloud to be registered; The point cloud filling process for the target registration point cloud specifically includes: Obtain the difference between the number of point clouds in the target registration point cloud and the number of point clouds in the point cloud to be registered; Randomly select a target registration point cloud with the same number of points as the difference from the target registration point cloud as the second filling point cloud; The second filled point cloud is added to the target registration point cloud.

5. A terminal device, characterized in that, The terminal device includes a memory, a processor, and a program for point cloud registration using random coefficients stored in the memory and executable on the processor. When the processor executes the program for point cloud registration using random coefficients, it implements the steps of the method for point cloud registration using random coefficients as described in any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for point cloud registration using random coefficients. When the program for point cloud registration using random coefficients is executed by a processor, it implements the steps of the method for point cloud registration using random coefficients as described in any one of claims 1-3.

Images