Global registration method and system for three-dimensional model point cloud data

Abstract

The present invention provides a global registration method and system for three-dimensional model point cloud data, the method comprising extracting point cloud data whose height is greater than the corresponding base height in the current single-frame input point cloud as point cloud data of interest; dividing the point cloud data of interest into multiple block point clouds based on a preset evaluation model, and registering the multiple block point clouds with the preset evaluation model to construct a new evaluation model; determining multiple block point clouds corresponding to the next single-frame input point cloud based on the new evaluation model; iterating the transformation matrix of the multiple block point clouds of each single-frame input point cloud, and converting the point cloud pose based on the transformation matrix formed after the iteration to output a converted point cloud set. The present invention solves the problem that registering three-dimensional point cloud data by rigid body transformation will introduce movement errors and cause a decrease in registration accuracy, or registering three-dimensional point cloud data by non-rigid body transformation will cause overall point cloud distortion, resulting in data deformation and thus low scanning accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of projectors, and in particular to a global registration method and system for three-dimensional model point cloud data. Background Art

[0002] Existing global registration methods for 3D model data usually iteratively adjust the spatial pose of multiple frame models to eliminate the stratification phenomenon of overlapping parts between models and minimize model splicing errors.

[0003] For example, the Chinese invention patent with publication number CN115984338A discloses a "global registration method based on spherical harmonic transformation and iterative optimization", which realizes the global registration of three-dimensional point clouds through a registration method based on spherical harmonic transformation and iterative optimization. In actual use, there is a situation where a single or multiple models are fixed by non-rigid connection methods such as blue glue. In the process of collecting three-dimensional point cloud data of models at different angles, each model will move relative to each other. At this time, the relative movement error will be introduced into the global registration through the rigid body transformation, resulting in a decrease in the registration accuracy, affecting the subsequent model data processing and use. For another example, the Chinese invention patent with publication number CN107133978A discloses a "three-dimensional non-rigid point cloud registration method based on consistent point drift algorithm", which performs non-rigid point cloud registration through a consistent point drift algorithm. Although non-rigid transformation can reduce the error caused by the relative movement of the model in data acquisition to a certain extent, it will cause overall point cloud distortion, causing deformation of the three-dimensional model data and reducing the scanning accuracy of the scanned object.

[0004] Therefore, it is necessary to improve the above scheme to solve the above technical problems. Summary of the invention

[0005] The purpose of the present invention is to provide a global registration method and system for three-dimensional model point cloud data, so as to solve the problem in the prior art that registering three-dimensional point cloud data by rigid transformation will introduce motion errors and cause a decrease in registration accuracy, or registering three-dimensional point cloud data by non-rigid transformation will cause overall point cloud distortion, resulting in data deformation and thus low scanning accuracy.

[0006] To achieve the above object, in a first aspect, a global registration method for three-dimensional model point cloud data is provided, the method comprising:

[0007] Determine the base height of each single-frame input point cloud frame by frame, so as to extract point cloud data whose height is greater than the corresponding base height in the current single-frame input point cloud as point cloud data of interest;

[0008] Based on a preset evaluation model, the point cloud data of interest is divided into a plurality of block point clouds, and the plurality of block point clouds are registered with the preset evaluation model to construct a new evaluation model;

[0009] Based on the new evaluation model, the point cloud data of interest determined based on the next single frame input point cloud is evaluated to determine a plurality of block point clouds corresponding to the next single frame input point cloud;

[0010] The transformation matrices of multiple block point clouds corresponding to each single frame input point cloud are iterated, and the point cloud pose is transformed based on the transformation matrix formed after the iteration to output a transformed point cloud set.

[0011] As a further improvement of the present invention, the base height of each single frame input point cloud is determined frame by frame, including:

[0012] Perform grid sampling on each single frame input point cloud to obtain the corresponding sparse point cloud;

[0013] The k-neighborhood variance of each point in the sparse point cloud in the corresponding single-frame input point cloud is determined, so as to determine the base height of each single-frame input point cloud based on the k-neighborhood variance.

[0014] As a further improvement of the present invention, the base height of each single frame input point cloud is determined based on the k-neighborhood variance, including:

[0015] Count the variance histogram of the height of each point whose k-neighborhood variance is less than the preset variance in the sparse point cloud corresponding to the single-frame input point cloud;

[0016] The upper limit of the height segment with the highest frequency in the variance histogram is taken as the base height of the corresponding single-frame input point cloud.

[0017] As a further improvement of the present invention, the point cloud data of interest is segmented into a plurality of block point clouds based on a preset evaluation model, including:

[0018] Performing error evaluation on the point cloud data of interest based on a preset evaluation model;

[0019] Based on the comparison result of the point cloud distance of each point in the point cloud data of interest and the preset error value, the point cloud data of interest is divided into a plurality of block point clouds.

[0020] As a further improvement of the present invention, based on the comparison result between the point cloud distance of each point in the point cloud data of interest and the preset error value, the point cloud data of interest is divided into a plurality of block point clouds, including:

[0021] A point cloud subset of the point cloud data of interest whose point cloud distance is less than a preset error value is used as the first block point cloud;

[0022] Repeat the error evaluation of the remaining point cloud data based on the preset evaluation model, so as to use the point cloud subset whose point cloud distance in the remaining point cloud data is less than the preset error value as the second block point cloud, until the remaining point cloud data is empty;

[0023] The remaining point cloud data is point cloud data of the point cloud data of interest excluding a point cloud subset whose point cloud distance is less than a preset error value.

[0024] As a further improvement of the present invention, the operation of performing error evaluation on the point cloud data of interest based on a preset evaluation model is specifically as follows:

[0025] The distance from each point in the point cloud data of interest to the nearest neighbor point of the preset evaluation model is taken as the point cloud distance.

[0026] In a second aspect, a global registration system for three-dimensional model point cloud data is provided, comprising:

[0027] The point cloud data determination unit of interest is used to determine the base height of each single frame input point cloud frame by frame, so as to extract the point cloud data whose height is greater than the corresponding base height in the current single frame input point cloud as the point cloud data of interest;

[0028] A data segmentation unit, configured to segment the point cloud data of interest into a plurality of block point clouds based on a preset evaluation model, and to evaluate the point cloud data of interest determined based on the next single frame input point cloud based on a new evaluation model to determine a plurality of block point clouds corresponding to the next single frame input point cloud;

[0029] An evaluation model construction unit, used for registering the plurality of block point clouds with the preset evaluation model to construct a new evaluation model;

[0030] An iteration unit, used for iterating the transformation matrices of multiple block point clouds corresponding to each single frame input point cloud; and

[0031] The data output unit is used to transform the point cloud pose based on the transformation matrix formed after iteration to output the transformed point cloud set.

[0032] As a further improvement of the present invention, it also includes:

[0033] A data sampling unit is used to perform grid sampling on each single frame input point cloud to obtain a corresponding sparse point cloud;

[0034] The base height determination unit is used to determine the k-neighborhood variance of each point in the sparse point cloud in the corresponding single-frame input point cloud, so as to determine the base height of each single-frame input point cloud based on the k-neighborhood variance.

[0035] As a further improvement of the present invention, it also includes:

[0036] A statistical unit, used to count the variance histogram of the height of each point whose k-neighborhood variance is less than a preset variance in the sparse point cloud corresponding to the single-frame input point cloud;

[0037] The base height determination unit is used to use the upper limit of the height segment with the highest frequency in the variance histogram as the base height of the corresponding single-frame input point cloud.

[0038] As a further improvement of the present invention, it also includes:

[0039] An error evaluation unit, used for performing error evaluation on the point cloud data of interest based on a preset evaluation model;

[0040] The data segmentation unit is used to segment the point cloud data of interest into a plurality of block point clouds based on a comparison result between a point cloud distance of each point in the point cloud data of interest and a preset error value.

[0041] In a third aspect, the present invention provides a terminal device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method described in the first aspect.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0043] The beneficial effects of the present invention are:

[0044] The global registration method of the three-dimensional model point cloud data of the present invention first removes the base parts with poor reconstruction quality and non-essential in the point cloud data through the base removal algorithm, and then segments the point cloud data of interest determined based on each single-frame input point cloud through the evaluation model (preset evaluation model and constructed new evaluation model) to release the rigid connection between the scanned parts that shake during the scanning process and release the error caused by the relative movement of the scanned parts, and then iterates the transformation matrix of the multiple block point clouds corresponding to each single-frame input point cloud through the iterative optimization algorithm to minimize the distance between each block point cloud, achieve the purpose of accurate registration, and thus improve the scanning accuracy of the model. In this way, it solves the problem in the prior art that registering three-dimensional point cloud data through rigid body transformation will introduce movement errors and cause a decrease in registration accuracy, or registering three-dimensional point cloud data through non-rigid body transformation will cause overall point cloud distortion, resulting in data deformation and low scanning accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic flow chart of a global registration method for three-dimensional model point cloud data according to an embodiment of the present invention;

[0046] Figure 2 A schematic flow chart of a global registration method for three-dimensional model point cloud data according to another embodiment of the present invention;

[0047] Figure 3 A schematic flow chart of a global registration method for three-dimensional model point cloud data according to another embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the effect after the base is removed;

[0049] Figure 5 A schematic flow chart of a global registration method for three-dimensional model point cloud data according to another embodiment of the present invention;

[0050] Figure 6 A schematic flow chart of a global registration method for three-dimensional model point cloud data according to another embodiment of the present invention;

[0051] Figure 7 This is a schematic diagram of the effect of dividing the point cloud into blocks;

[0052] Figure 8 This is a schematic diagram of the effect of the new evaluation model;

[0053] Fig. 9 This is the final effect of global registration;

[0054] Fig.10 A schematic flow chart of a global registration method for three-dimensional model point cloud data according to a specific embodiment of the present invention;

[0055] Fig.11 A schematic structural block diagram of a global registration system for three-dimensional model point cloud data according to an embodiment of the present invention;

[0056] Fig.12 A topological structure diagram of a computer-readable storage medium disclosed in the present invention. DETAILED DESCRIPTION

[0057] The present invention is described in detail below in conjunction with the various embodiments shown in the accompanying drawings, but it should be noted that these embodiments are not limitations of the present invention, and any equivalent transformations or substitutions in functions, methods, or structures made by ordinary technicians in the field based on these embodiments are all within the scope of protection of the present invention.

[0058] The technical solutions provided by various embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0059] Embodiment 1:

[0060] Figure 1The present invention is a schematic flow chart of a global registration method for three-dimensional model point cloud data (hereinafter referred to as "global registration method" or "registration method" or "method"), which is used to solve the problem that the registration of three-dimensional point cloud data by rigid body transformation will introduce movement error and cause the registration accuracy to decrease, or the registration of three-dimensional point cloud data by non-rigid body transformation will cause the overall point cloud to be distorted, resulting in data deformation and low scanning accuracy. The method includes:

[0061] Step 102: Determine the base height of each single-frame input point cloud P frame by frame, so as to extract the point cloud data whose height is greater than the corresponding base height in the current single-frame input point cloud P as the point cloud data of interest P. 0 .

[0062] Among them, Figure 2 As shown, the specific operations of step 102 include:

[0063] Step 202: Perform grid sampling on each single frame input point cloud P to obtain the corresponding sparse point cloud P 1 * .

[0064] Step 204: Determine the sparse point cloud P * The k-neighborhood variance convP of each point in the corresponding single-frame input point cloud P * , to determine the base height h of each single-frame input point cloud based on the k-neighborhood variance. The k-neighborhood variance convP of the single-frame input point cloud P * As shown in Formula 1:

[0065]

[0066] Among them, neiP * Indicates P * The midpoint is in the k-neighborhood point set of the input point cloud P, P i Represents the i-th point cloud of the k-neighborhood point set of the single-frame input point cloud P, It represents the mean of the k-neighborhood point set. k is a constant that can be set by the user as required.

[0067] like Figure 3 As shown, the specific operations of step 204 include:

[0068] Step 302: Count the variance histogram of the height of each point in the sparse point cloud corresponding to the single-frame input point cloud whose k-neighborhood variance is less than the preset variance.

[0069] Step 304. The upper limit of the height segment with the highest frequency in the variance histogram is used as the base height h of the corresponding single-frame input point cloud. The effect after removing the base from the input point cloud data is as follows: Figure 4 shown.

[0070] Step 104: Segment the point cloud data of interest into a plurality of block point clouds based on the preset evaluation model, and align the plurality of block point clouds with the preset evaluation model to construct a new evaluation model.

[0071] like Figure 5 As shown, the step of "segmenting the point cloud data of interest into a plurality of block point clouds based on a preset evaluation model" in step 104 specifically includes:

[0072] Step 502. Based on the preset evaluation model P E For the point cloud data P 0 Perform error assessment.

[0073] Specifically, the error evaluation method is: the distance from each point in the point cloud data of interest to the nearest neighbor point of the preset evaluation model is used as the point cloud distance.

[0074] Step 504: based on the comparison result of the point cloud distance of each point in the point cloud data of interest and the preset error value, the point cloud data of interest is divided into a plurality of block point clouds.

[0075] Specifically, if Figure 6 As shown, the specific operations of step 504 include:

[0076] Step 602: The point cloud data P 0 The point cloud subset whose midpoint cloud distance is less than the preset error value is taken as the first block point cloud P 0 1 .

[0077] Step 604: Repeat the process based on the preset evaluation model P E The remaining point cloud data is evaluated for errors, so that the point cloud subset whose point cloud distance is less than the preset error value in the remaining point cloud data is used as the second block point cloud P 0 2 , until the remaining point cloud data is empty. The remaining point cloud data is the point cloud data of the point cloud data of interest excluding the point cloud subset whose point cloud distance is less than the preset error value.

[0078] It should be understood that after obtaining the second block point, the error evaluation is performed based on the preset evaluation model on the point cloud data of interest except for the point cloud subset whose point cloud distance is less than the preset error value (the point cloud subset whose point cloud distance is less than the preset error value is the first block point cloud and the second block point cloud) to segment the third block point cloud P 0 3 , and so on to get s block point clouds P 0 3 , until the remaining point cloud data is empty. Finally, the first block point cloud P 0 1 , the second block point cloud P0 2 , the third block point cloud P 0 3 etc. to form multiple block point clouds P 0 '. That is, P 0 '=P 0 1 ∪P 0 2 ∪P 0 3 …∪P 0 s It should be noted that if the preset evaluation model P E If it is empty, the block point cloud is the point cloud data of interest P 0 Among them, the effects of multiple block point clouds are as follows Figure 7 shown.

[0079] In step 104, “registering the multiple block point clouds with the preset evaluation model to construct a new evaluation model” specifically includes: registering the multiple block point clouds P 0 'Perform ICP (Iterative Closest Point) registration with the preset evaluation model, and then perform the registration on the point cloud set P 0 '∪P E Perform grid sampling to form a newly constructed evaluation model P E ', its effect is as follows Figure 8 As shown, this paves the way for evaluating subsequent single-frame input point cloud data.

[0080] Step 106: Evaluate the point cloud data of interest determined based on the next single frame input point cloud based on the new evaluation model to determine a plurality of block point clouds corresponding to the next single frame input point cloud.

[0081] That is, for the input m-frame single-frame point cloud {P m Repeat the above steps 102 to 106 frame by frame to obtain m block point clouds {P 0m '}.

[0082] Step 108. Iterate the transformation matrices of the multiple block point clouds corresponding to each single frame input point cloud, and transform the point cloud pose based on the transformation matrix formed after the iteration to output a transformed point cloud set.

[0083] The Levenberg-Marquardt algorithm is used to iteratively optimize n block point clouds P 0n ', that is, the transformation matrix corresponding to the minimum point cloud spacing is calculated according to Formula 2.

[0084]

[0085] Among them, T * Represents the set of final transformation matrices obtained after iteration (i.e., T after iteration i and T j The collection of C ij P 0 i and P 0 j The neighbor point set of T i and T j They are divided into block point clouds P 0 i and P 0 j The transformation matrix is ​​the pose. In the initial iteration, T i and T j Set an initial value T i0 , T j0 Enter the iteration to iteratively output the block point cloud P 0 i and P 0 j The transformation matrix T i and T j .

[0086] For m block point clouds {P 0m '}Use the transformation matrix obtained by iteration based on formula 2 to transform the point cloud set, so as to output the transformed point cloud set. The effect after global transformation is as follows Fig. 9 shown.

[0087] It can be understood that the global registration method of the three-dimensional model point cloud data of this embodiment first removes the base parts with poor reconstruction quality and non-essential in the point cloud data through the base removal algorithm, and then segments the point cloud data of interest determined based on each single-frame input point cloud through the evaluation model (the preset evaluation model and the constructed new evaluation model) to release the rigid connection between the scanned parts that shake during the scanning process and release the errors caused by the relative movement of the scanned parts. Then, the transformation matrix of the multiple block point clouds corresponding to each single-frame input point cloud is iterated through the iterative optimization algorithm to minimize the distance between each block point cloud, so as to achieve the purpose of accurate registration, thereby improving the scanning accuracy of the model. In this way, the method of this embodiment solves the problem in the prior art that the registration of three-dimensional point cloud data by rigid transformation will introduce movement errors and cause a decrease in registration accuracy, or the registration of three-dimensional point cloud data by non-rigid transformation will cause overall point cloud distortion, resulting in data deformation and low scanning accuracy.

[0088] like Fig.10As shown, in a specific embodiment, the global registration method of the three-dimensional model point cloud data of this embodiment mainly includes:

[0089] Step 1001. Grid sampling is performed on each single frame input point cloud P to obtain the corresponding sparse point cloud P 1 * .

[0090] Step 1002. Calculate the sparse point cloud P according to formula 1 * The k-neighborhood variance convP of each point in the corresponding single-frame input point cloud P * .

[0091] Step 1003: Count the variance histogram of the height of each point in the sparse point cloud corresponding to the single-frame input point cloud whose k-neighborhood variance is less than the preset variance.

[0092] Step 1004: Output the upper limit of the height segment with the highest frequency in the variance histogram, that is, use the upper limit of the height segment with the highest frequency in the variance histogram as the base height h of the corresponding single-frame input point cloud.

[0093] Step 1005. For the point cloud data P of interest 0 Specifically, each point in the point cloud data of interest is converted to the preset evaluation model P E The distance between the nearest neighbor points is taken as the point cloud distance.

[0094] Step 1006: based on the comparison result between the point cloud distance of each point in the point cloud data of interest and the preset error value, the point cloud data of interest is divided into a plurality of block point clouds P 0 ', and determine multiple block point clouds corresponding to other single-frame point clouds based on the new evaluation model.

[0095] Among them, multiple block point clouds P 0 'Perform ICP (Iterative Closest Point) registration with the preset evaluation model, and then perform the registration on the point cloud set P 0 '∪P E Perform grid sampling to form a newly constructed evaluation model P E ', thus paving the way for evaluating the subsequent single-frame input point cloud data. For the input m-frame single-frame point cloud {P m Repeat the above steps 1001 to 1006 frame by frame to obtain m block point clouds {P 0m '}.

[0096] Step 1007: Globally optimize the multiple block point clouds corresponding to each single frame input point cloud, that is, iterate the transformation matrix of the multiple block point clouds corresponding to each single frame input point cloud according to Formula 2, and transform the point cloud pose based on the transformation matrix formed after the iteration.

[0097] Step 1008: Output the optimized point cloud set.

[0098] In this embodiment, the global registration method of the point cloud data of the three-dimensional model first removes the base parts with poor reconstruction quality and unnecessary base parts in the point cloud data through the base removal algorithm, and then segments the point cloud data of interest determined based on each single-frame input point cloud through the evaluation model to release the rigid connection between the scanned parts that shake during the scanning process and release the errors caused by the relative movement of the scanned parts. Then, the transformation matrices of the multiple block point clouds corresponding to each single-frame input point cloud are iterated through the iterative optimization algorithm to minimize the distance between each block point cloud, thereby achieving the purpose of precise registration and improving the scanning accuracy of the model.

[0099] Embodiment 2:

[0100] like Fig.11 As shown, the present embodiment further provides a global registration system 1100 for three-dimensional model point cloud data, including: an interest point cloud data determination unit 111, used to determine the base height of each single frame input point cloud frame by frame, so as to extract the point cloud data whose height is greater than the corresponding base height in the current single frame input point cloud as the interest point cloud data; a data segmentation unit 112, used to segment the interest point cloud data into a plurality of block point clouds based on a preset evaluation model, and to evaluate the interest point cloud data determined based on the next single frame input point cloud based on a new evaluation model, so as to determine the plurality of block point clouds corresponding to the next single frame input point cloud; an evaluation model construction unit 113, used to align the plurality of block point clouds with the preset evaluation model, so as to construct a new evaluation model; an iteration unit 114, used to iterate the transformation matrix of the plurality of block point clouds corresponding to each single frame input point cloud; and a data output unit 115, used to convert the point cloud pose based on the transformation matrix formed after the iteration, so as to output the converted point cloud set.

[0101] Furthermore, the system of this embodiment also includes: a data sampling unit 116, which is used to perform grid sampling on each single-frame input point cloud to obtain a corresponding sparse point cloud; a base height determination unit 118, which is used to determine the k-neighborhood variance of each point in the sparse point cloud corresponding to the single-frame input point cloud, so as to determine the base height of each single-frame input point cloud based on the k-neighborhood variance. Among them, the system of this embodiment may also include a statistical unit 117, through which the statistical unit 117 counts the variance histogram of the height of each point whose k-neighborhood variance is less than the preset variance in the sparse point cloud corresponding to the single-frame input point cloud; the base height determination unit 117 uses the upper limit of the height segment with the highest frequency in the variance histogram as the base height of the corresponding single-frame input point cloud. The system of this embodiment may also include an error evaluation unit 119, which is used to perform error evaluation on the point cloud data of interest based on a preset evaluation model; the data segmentation unit 112 segments the point cloud data of interest into multiple block point clouds according to the comparison result of the point cloud distance of each point in the point cloud data of interest and the preset error value. Specifically, the data segmentation unit 112 can also be used to take a point cloud subset in the point cloud data of interest whose point cloud distance is less than a preset error value as the first block point cloud, and repeatedly perform error evaluation on the remaining point cloud data based on a preset evaluation model, so as to take a point cloud subset in the remaining point cloud data whose point cloud distance is less than the preset error value as the second block point cloud, until the remaining point cloud data is empty; wherein the remaining point cloud data is the point cloud data in the point cloud data of interest excluding the point cloud subset whose point cloud distance is less than the preset error value.

[0102] It should be understood that the global registration system 1100 of the three-dimensional model point cloud data of the present embodiment first removes the base parts with poor reconstruction quality and non-essential in the point cloud data by the base removal algorithm through the interest point cloud data determination unit 111, and then segments the interest point cloud data determined based on each single-frame input point cloud according to the evaluation model (the preset evaluation model and the constructed new evaluation model, the new evaluation model is constructed by the evaluation model construction unit 113 by aligning multiple block point clouds with the preset evaluation model to obtain a new evaluation model) through the data segmentation unit 112, so as to release the rigid connection between the scanned parts that shake during the scanning process and release the errors caused by the relative movement of the scanned parts. Then, the transformation matrix of the multiple block point clouds corresponding to each single-frame input point cloud is iterated by the iteration unit 114 based on the iterative optimization algorithm to minimize the distance between each block point cloud, and finally the converted point cloud set is output through the data output unit 115 to achieve the purpose of precise registration, thereby improving the scanning accuracy of the model. In this way, the system of this embodiment solves the problem in the prior art that aligning three-dimensional point cloud data through rigid transformation will introduce motion errors and cause a decrease in alignment accuracy, or aligning three-dimensional point cloud data through non-rigid transformation will cause overall point cloud distortion, resulting in data deformation and thus low scanning accuracy.

[0103] It should be noted that the global registration system of the three-dimensional model point cloud data in this embodiment has the same technical solutions as those in the first embodiment, and please refer to the first embodiment, which will not be repeated here.

[0104] Embodiment three:

[0105] The embodiment of the present invention further provides a terminal device, which may include a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor to implement the above Figure 1 The various processes of the embodiment of the global registration method for three-dimensional model point cloud data shown in the figure can achieve the same technical effect, and will not be described again here to avoid repetition.

[0106] Embodiment 4:

[0107] Combination Fig.12 As shown, this embodiment also discloses a specific implementation of a computer-readable storage medium 1200. The computer-readable storage medium 1200 can be configured in whole or in part in a physical computer, server, cluster server or data center.

[0108] In this embodiment, the computer-readable storage medium 1200 stores computer program instructions 1201. When the computer program instructions 1201 are read and executed by a processor 1202, the steps in the global registration method of three-dimensional model point cloud data disclosed in the first embodiment are executed.

[0109] Optionally, the computer-readable storage medium 1200 can be configured as a server, and the server runs on a physical device for building a private cloud, a hybrid cloud, or a public cloud. At the same time, the computer-readable storage medium 1200 can also be configured as a random access memory (Random Access Memory, RAM), a read-only memory (Read Only Memory, ROM), a programmable read-only memory (Programmable Read-Only Memory, PROM), an erasable read-only memory (ErasableProgrammable Read-Only Memory, EPROM), an electrically erasable read-only memory (Electric ErasableProgrammable Read-Only Memory, EEPROM), etc.

[0110] The computer-readable storage medium 1200 is used to store programs, and the processor 1202 executes the global registration method of three-dimensional model point cloud data disclosed in the first embodiment after receiving the execution instruction.

[0111] At the same time, the processor 1202 disclosed in this embodiment may be an integrated circuit chip with signal processing capabilities. The processor 1202 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention may be implemented or executed. The general-purpose processor may be a microprocessor or the general-purpose processor may also be any conventional processor.

[0112] For the technical solutions of the computer-readable storage medium 1200 disclosed in this embodiment that are the same as those in Embodiment 1 and / or Embodiment 2, please refer to Embodiment 1 and / or Embodiment 2, which will not be repeated here.

[0113] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation methods or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

[0114] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0115] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A global registration method for three-dimensional model point cloud data, characterized in that: The method comprises: Determine the base height of each single-frame input point cloud frame by frame, so as to extract point cloud data whose height is greater than the corresponding base height in the current single-frame input point cloud as point cloud data of interest; Based on a preset evaluation model, the point cloud data of interest is divided into a plurality of block point clouds, and the plurality of block point clouds are registered with the preset evaluation model to construct a new evaluation model; Based on the new evaluation model, the point cloud data of interest determined based on the next single frame input point cloud is evaluated to determine a plurality of block point clouds corresponding to the next single frame input point cloud; Iterate the transformation matrices of multiple block point clouds corresponding to each single frame input point cloud, and transform the point cloud pose based on the transformation matrix formed after the iteration to output a transformed point cloud set; Wherein, the point cloud data of interest is divided into a plurality of block point clouds based on a preset evaluation model, including: Performing error evaluation on the point cloud data of interest based on a preset evaluation model; Based on the comparison result of the point cloud distance of each point in the point cloud data of interest and the preset error value, the point cloud data of interest is divided into multiple block point clouds, wherein the distance from each point in the point cloud data of interest to the nearest neighbor point of the preset evaluation model is taken as the point cloud distance.

2. The global registration method according to claim 1, characterized in that: Determine the base height of each single frame input point cloud frame by frame, including: Perform grid sampling on each single frame input point cloud to obtain the corresponding sparse point cloud; The k-neighborhood variance of each point in the sparse point cloud in the corresponding single-frame input point cloud is determined, so as to determine the base height of each single-frame input point cloud based on the k-neighborhood variance.

3. The global registration method according to claim 2, characterized in that: The base height of each single frame input point cloud is determined based on the k-neighborhood variance, including: Count the variance histogram of the height of each point whose k-neighborhood variance is less than the preset variance in the sparse point cloud corresponding to the single-frame input point cloud; The upper limit of the height segment with the highest frequency in the variance histogram is taken as the base height of the corresponding single-frame input point cloud.

4. The global registration method according to claim 1, characterized in that: Based on the comparison result between the point cloud distance of each point in the point cloud data of interest and the preset error value, the point cloud data of interest is divided into a plurality of block point clouds, including: A point cloud subset of the point cloud data of interest whose point cloud distance is less than a preset error value is used as the first block point cloud; Repeat the error evaluation of the remaining point cloud data based on the preset evaluation model, so as to use the point cloud subset whose point cloud distance in the remaining point cloud data is less than the preset error value as the second block point cloud, until the remaining point cloud data is empty; The remaining point cloud data is point cloud data of the point cloud data of interest excluding a point cloud subset whose point cloud distance is less than a preset error value.

5. A global registration system for three-dimensional model point cloud data, characterized in that: include: The point cloud data determination unit of interest is used to determine the base height of each single frame input point cloud frame by frame, so as to extract the point cloud data whose height is greater than the corresponding base height in the current single frame input point cloud as the point cloud data of interest; A data segmentation unit, configured to segment the point cloud data of interest into a plurality of block point clouds based on a preset evaluation model, and to evaluate the point cloud data of interest determined based on the next single frame input point cloud based on a new evaluation model to determine a plurality of block point clouds corresponding to the next single frame input point cloud; An evaluation model construction unit, used for registering the plurality of block point clouds with the preset evaluation model to construct a new evaluation model; An iteration unit, used for iterating the transformation matrices of multiple block point clouds corresponding to each single frame input point cloud; as well as A data output unit, used for converting the point cloud pose based on the transformation matrix formed after iteration, so as to output a converted point cloud set; An error evaluation unit, used for performing error evaluation on the point cloud data of interest based on a preset evaluation model; The data segmentation unit is used to segment the point cloud data of interest into multiple block point clouds based on the comparison result of the point cloud distance of each point in the point cloud data of interest and the preset error value, wherein the distance from each point in the point cloud data of interest to the nearest neighbor point of the preset evaluation model is used as the point cloud distance.

6. The global registration system according to claim 5, characterized in that: Also includes: A data sampling unit is used to perform grid sampling on each single frame input point cloud to obtain a corresponding sparse point cloud; The base height determination unit is used to determine the k-neighborhood variance of each point in the sparse point cloud in the corresponding single-frame input point cloud, so as to determine the base height of each single-frame input point cloud based on the k-neighborhood variance.

7. The global registration system according to claim 6, characterized in that: Also includes: A statistical unit, used to count the variance histogram of the height of each point whose k-neighborhood variance is less than a preset variance in the sparse point cloud corresponding to the single-frame input point cloud; The base height determination unit is used to use the upper limit of the height segment with the highest frequency in the variance histogram as the base height of the corresponding single-frame input point cloud.

Images