CN117291960A - Moon surface point cloud global registration method and equipment based on overlap degree classification - Google Patents

Abstract

The invention relates to a global registration method and equipment for moon surface point clouds based on overlap classification, wherein the method comprises the following steps: acquiring a plurality of overlapped or partially overlapped point clouds, extracting the boundary of each point cloud, calculating the overlapping degree between the point clouds, and dividing the point clouds into an overlapped point cloud set and an isolated point cloud set based on the overlapping degree and a preset threshold value; acquiring a reference digital terrain model of a moon surface, obtaining reference point clouds through dispersion, and carrying out density consistency processing on each single-view point cloud in the overlapped point cloud set and the isolated point cloud set; performing iterative registration with the reference point cloud based on global generalized general analysis and iterative nearest point algorithm aiming at the overlapped point cloud set; registering the isolated point cloud set with the reference point cloud based on an iterative nearest point algorithm; and (3) completing global point cloud registration to obtain the multi-view cloud after global registration. Compared with the prior art, the method can generate the fusion terrain model with higher quality.

Description

Moon surface point cloud global registration method and equipment based on overlap degree classification

Technical Field

The invention relates to the field of point cloud registration and fusion, in particular to a moon surface point cloud global registration method and equipment based on overlap classification.

Background

Because of the rapid progress of photogrammetry and computer vision technology, three-dimensional point clouds generated by dense image matching have become an important data source for lunar topographic mapping due to their accuracy, reliability and lower cost. The method is influenced by factors such as the attitude precision of the camera outside, the dense matching precision and the like, and the phenomenon of layering, staggered layers and the like still exist among the point clouds, so that great difficulty is caused to the subsequent multi-view cloud fusion. In order to realize the large-scale drawing requirement of the lunar surface, global registration is required to be carried out on the multi-view point clouds, the position consistency among the point clouds is improved, and a proper weighting strategy is adopted in the fusion stage to obtain a fusion terrain model with higher quality.

Global registration of point clouds is a more complex and challenging research problem than two-view point cloud registration, and is mainly divided into sequential registration and global registration. The sequential registration method involves sequentially aligning two overlapping views each time, and global registration mainly solves the problem that accumulated registration errors generated in the sequential registration process uniformly distribute the errors to all overlapping views.

The chinese application publication No. CN 110009667A provides a multi-view cloud global registration method based on the rodgers transformation, firstly, on the basis of obtaining a plurality of multi-view clouds which are taken from different view angles and partially overlap each other between the blocks, searching a matching point pair between any two point clouds with overlapping relation, calculating a corresponding rotation transformation matrix and translation transformation vector of each point cloud compared with the corresponding rotation transformation matrix and translation transformation vector, then converting all the calculated rotation transformation matrix into rotation transformation vector by using the rodgers transformation, merging the rotation transformation vector and translation transformation vector as observation values into the multi-view cloud global optimization model, and obtaining each optimal transformation matrix of the multi-view scanning surface point clouds through repeated iterative adjustment calculation, thereby completing the integral accurate registration of the multi-view scanning surface point clouds.

The application can realize global registration of multi-view clouds with a certain overlapping relationship, but does not perform classification processing on different overlapping degrees, so that the registration accuracy is not ideal.

Disclosure of Invention

The invention aims to overcome the defects of the prior art and provide a global registration method and equipment for moon surface point cloud based on overlap classification, so as to solve or partially solve the problem of non-ideal registration precision of point cloud.

The aim of the invention can be achieved by the following technical scheme:

the invention provides a global registration method of a moon surface point cloud based on overlap degree classification, which comprises the following steps:

acquiring a plurality of overlapped or partially overlapped point clouds, extracting the boundary of each point cloud, calculating the overlapping degree between the point clouds, and dividing the point clouds into an overlapped point cloud set and an isolated point cloud set based on the overlapping degree and a preset threshold value;

acquiring a reference digital terrain model of a moon surface, obtaining reference point clouds through dispersion, and carrying out density consistency processing on each single-view point cloud in the overlapped point cloud set and the isolated point cloud set;

performing iterative registration with the reference point cloud based on global generalized general analysis and iterative nearest point algorithm aiming at the overlapped point cloud set;

registering the isolated point cloud set with the reference point cloud based on an iterative nearest point algorithm;

and (3) completing global point cloud registration to obtain the multi-view cloud after global registration.

As a preferred technical solution, the calculating process of the overlapping degree between the point clouds includes:

and projecting the point cloud on a two-dimensional plane, acquiring a boundary point set of the point cloud based on an alhpa shapes algorithm, simplifying the boundary point set based on a convex hull algorithm, and calculating the overlapping degree between the point clouds through a geos library.

As a preferable embodiment, the density matching process includes:

and enabling the density of each single-view point cloud in the overlapped point cloud set and the isolated point cloud set to be consistent with the density of the reference point cloud through voxel filtering.

As a preferred technical solution, the process of registering the overlapping point cloud sets includes:

searching the nearest matching relation in the rest point clouds aiming at each point in each point cloud, calculating a rotation matrix and a translation matrix of each point cloud based on the nearest matching relation and a preset objective function, updating the point clouds, and repeatedly executing the steps until a preset global convergence condition is met, thereby completing registration.

As a preferable technical solution, the global convergence condition is:

the iteration times exceed a preset value, or the global registration errors before and after global registration are smaller than the preset value, wherein the global registration errors are based on the average value of the corresponding point distances from the point cloud to the centroid cloud, and the centroid cloud is calculated based on the independent subset of each nearest matching relation.

As a preferable technical scheme, the objective function is:

wherein m is the number of overlapping point clouds, A ₁ ,A ₂ ,…,A _m Respectively representing m point sets, p is the number of point clouds in the point cloud set, and letT is a rotation matrix, j is a p×1 unit vector, T is a translation matrix, and p×p matrix M _i Representing the visibility of the point set, 1 being visible, 0 being absent, < ->

As a preferred technical scheme, the method further comprises:

based on the multi-view cloud after global registration, a spatial grid fusion strategy based on mixed weighting of triangulation errors and distances is adopted to obtain fusion point cloud, and interpolation is carried out to obtain a fusion digital terrain model;

and carrying out gradient-based outlier detection and contour detection-based hole filling on the fused digital terrain model to finish fusion of the point cloud and the digital terrain model.

As a preferred technical solution, the process for obtaining the fused digital terrain model includes:

calculating the size parameter of a minimum bounding box of the point cloud based on extreme values on three coordinate axes of the multi-view cloud three-point set after global registration;

dividing the minimum bounding box into a plurality of space grids based on the set voxel small grid side length and numbering;

and calculating a space grid fusion value by adopting a fusion strategy of mixed weighting of the triangulation error and the distance to obtain the fused digital terrain model.

As an preferable technical scheme, the calculation of the spatial grid fusion value is realized by adopting the following formula:

wherein w is _i,t Represents the ith ^th Point-to-t ^th The weight in the grid is determined by the weight in the grid,representing the spatial grid fusion value, d _i,t And e _i,t Respectively representing Euclidean distance and triangulation error between ith point and tth grid center, alpha ₁ Sum alpha ₂ The weight values of the relative error and distance terms are triangulated, respectively.

In another aspect of the present invention, there is provided an electronic apparatus including: the system comprises one or more processors and a memory, wherein one or more programs are stored in the memory, and the one or more programs comprise instructions for executing the global registration method of the moon point cloud based on the overlapping degree classification.

Compared with the prior art, the invention has the following advantages:

(1) The point cloud registration accuracy is high: the method overcomes or partially overcomes the defect of low precision of multi-view cloud registration under the condition of complex point cloud coverage, combines global GPA+ICP adjustment with ICP registration based on reference point cloud, and realizes global registration of the multi-view cloud of the lunar surface by utilizing the global registration method of the lunar surface point cloud according to the overlapping degree classification by utilizing the two methods respectively, thereby improving the position consistency and absolute precision between the point clouds.

(2) The registering efficiency is high: the existing global registration method does not fully consider the low overlapping rate of the real photogrammetry point cloud, and cannot guarantee reliable registration accuracy for complex point cloud distribution conditions.

(3) The accuracy of the fused digital terrain model is high: the invention adopts a weighting strategy which comprehensively considers three-dimensional point triangulation errors and distances, is used for multi-view cloud fusion, effectively eliminates fine geometric inconsistencies still existing in the multi-view cloud, simultaneously generates a smooth terrain model, and simultaneously performs outlier detection and hole filling on the fused terrain model to ensure the integrity and precision of the terrain model.

Drawings

FIG. 1 is a flowchart of a global registration method for a moon surface point cloud based on overlap classification in an embodiment;

FIG. 2 is a schematic diagram of a point cloud coverage situation and a point cloud set distribution situation with a degree of overlap of more than 50% in the embodiment;

FIG. 3 is a schematic view of a reference DEM and a reference point cloud in an embodiment;

FIG. 4 is a schematic diagram of the point cloud density unification in the embodiment;

fig. 5 is a schematic diagram of distribution of the point clouds of the front and rear projection of gpa+icp global registration in the embodiment;

FIG. 6 is a schematic diagram of grade-based anomaly grid detection in an embodiment;

FIG. 7 is a schematic diagram of hole filling based on contour detection in an embodiment;

fig. 8 is a cross-sectional view of a DEM and mountain in an embodiment.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.

Example 1

In order to provide a global registration and fusion method of lunar surface point clouds classified according to overlapping degree, overcome uncertainty of registration errors caused by complex point cloud coverage conditions, and promote position consistency after global registration of multi-view point clouds, so as to generate a fused terrain model with higher quality, referring to fig. 1, the embodiment provides a global registration method of lunar surface point clouds based on overlapping degree classification, which comprises the following steps.

S1, obtaining all point cloud boundaries, and calculating the overlapping degree between the point clouds. The method specifically comprises the following substeps.

S11, projecting points on a two-dimensional plane;

s12, acquiring a point cloud boundary point set by using an alpha shapes algorithm;

s13, simplifying a point cloud boundary point set by using a convex hull algorithm;

s14, calculating the overlapping degree between the point clouds through a geos library.

The point cloud is a photogrammetric point cloud generated by a multi-view LROC NAC image of the lunar orbit device, and the total number of the point clouds is more than or equal to 3.

Referring to fig. 2, (a) calculates the overlapping degree between the point clouds of the Apollo17 landing area in the present disclosure for the coverage situation of the total 12 scenic point clouds. Obtaining a point cloud set a {4,8,11,12}, wherein the overlapping degree between the point clouds is more than 50%, and the point clouds have sufficient overlapping degree (as shown in (b) of fig. 2); ICP registration based on reference point clouds is adopted for the rest isolated point clouds, and point clouds 2, 3 and 5 are taken as examples.

S2, discretizing a reference digital terrain model (DEM) to obtain a reference point cloud.

Specifically, a three-dimensional point is generated for each pixel unit, and geographic coordinates and elevation information are given to the three-dimensional point, wherein the geographic coordinates are obtained through calculation of initial coordinate information in a DEM header file and ground resolutions in x and y directions, and the elevation information is a pixel value of each pixel.

According to the method, SLDEM2015 data (downloading address: https:// im brium. Mit. Edu/EXTRAS/SLDEM2015 /) is selected with reference to the DEM, the resolution is about 60m/pix, preprocessing is carried out on the SLDEM2015 data, the steps of altitude value reduction, projection and cutting are included, and the three-dimensional point cloud is obtained according to the discretization method. Fig. 3 (a) (b) is a reference DEM and a reference point cloud, respectively.

S3, processing the cloud images with the calculated overlapping degree in the S1 to obtain down-sampling point clouds with the same density as the reference point clouds.

Specifically, the density of the downsampled single-view point cloud is ensured to be consistent with that of the reference point cloud through voxel filtering, and the formula is as follows:

wherein the minimum value of the point cloud point set is X _min 、Y _min 、Z _min The voxel grid has a side length of cell, the t-th voxel grid is numbered (i, j, k), and t=i·l _x +j·l _y +k。

And obtaining downsampling parameters by voxel filtering and parameter adjustment: 60. fig. 4 (a) and (b) show the conditions before and after the point cloud downsampling, the red point is an SLDEM point cloud, the blue is a single-view point cloud, and the density of the two point clouds tends to be consistent intuitively.

And S4, registering the point cloud set with high overlapping degree by adopting global generalized Pu' S analysis and iterative closest point (GPA+ICP) algorithm, and evaluating global adjustment precision. The method comprises the following substeps.

S41, searching the nearest matching relation in the rest point cloud views for all points in each scenic point cloud.

Specifically, for the corresponding point obtained by nearest neighbor search of iterative nearest point (ICP) algorithm, a current point p is defined _i ∈A _i And p is as follows _j ∈A _j Are the nearest neighbors of each other (i.e. p _i At A _j The nearest neighbor point in (a) is p _j ，p _j At A _i The nearest neighbor point in (a) is p _i ) Is the "nearest matching relationship".

S42, calculating the mass center K for each independent subset of the nearest matching relations.

S43, solving a rotation matrix T of each point cloud by a minimized function ₁ ,T ₂ ,…,T _m Translation matrix t ₁ ,t ₂ ,…,t _m 。

Wherein the minimization objective function is:

where m is the number of point clouds (point set A ₁ ,A ₂ ,…,A _m ) P is the number of point sets, letT is a 3×3 rotation matrix, j is a p×1 unit vector, T is a 3×1 translation matrix, and p×p matrix M is introduced _i To indicate the visibility of the point set, 1 is visible, 0 is absent, let +.>

S44, the obtained transformation matrix acts on each scenic spot cloud to obtain updated point clouds.

S45, returning to the step S41, and iterating until the global convergence.

Specifically, the Point-to-Point ICP is adopted to register the two, and the maximum iteration times, the difference value between the two transformation matrixes and the mean square error value are respectively set as conditions for ICP iteration convergence.

After the iterative calculation is completed, the accuracy of the method is evaluated by calculating GPA+ICP global registration errors and absolute registration errors.

The single-view point clouds are respectively downsampled to 150 meters, and then the downsampled point clouds are used as input data for global adjustment, wherein the three-dimensional point clouds respectively comprise 16346, 12710, 9995 and 28206 three-dimensional points. The iteration termination condition set in the method is that (errAfter-errPrev)/errAfter <1e-10 or the iteration times reach 100 times, wherein the errAfter and the errPrev are global registration errors before and after global adjustment (refer to average values of corresponding point distances from each scenic point cloud to a centroid cloud) respectively, and the global adjustment process iterates 31 times to convergence. Table 1 shows RT transformation matrix obtained by global ICP adjustment of four-point cloud. Fig. 5 (a) and (b) are respectively global adjustment front and rear point cloud distribution conditions, so that the apparent elevation inconsistency between the point clouds before registration is visually seen, the point clouds 11 are obviously higher than the other three views, and good space self-consistency between the four-view point clouds can be seen after global registration.

TABLE 1 Point cloud RT matrix

And S5, registering the globally adjusted point cloud group to the reference point cloud by utilizing ICP, and directly registering the isolated point cloud with low overlapping degree to the reference point cloud.

The four scenic spot clouds (point cloud set a) subjected to global GPA+ICP registration are integrally registered to a reference point cloud, and specific configuration parameters of ICP are as follows: maximum number of iterations: 500; difference between the twice transformed matrices: 1e-10; mean square error: 1, solving an RT transformation matrix obtained by the solution is shown in a table 2; for the isolated point cloud with low overlapping degree with the adjacent point cloud, the isolated point cloud is directly registered to the reference point cloud, the ICP registration parameters are consistent with the above, the effects of a single Jing Peizhun of the point clouds 2, 3 and 5 are listed below, and the RT matrix is shown in table 2.

TABLE 2 Point cloud and isolated Point cloud RT matrix

And S6, evaluating the global point cloud registration accuracy by utilizing a plurality of checking methods.

Firstly, evaluating the global GPA and ICP registration accuracy, wherein the global GPA and ICP registration error (refer to the average value of the corresponding point distances from each scenic spot cloud to the centroid cloud) is reduced from 14.0388 meters to 7.04875 meters, and secondly, the overall registration accuracy is reflected by evaluating the error between each pair of point clouds before and after global registration, and particularly, the table 3 shows that no overlapping exists between the point clouds or the overlapping area is very small and indicated by "/". After global registration, the mean value of the registration errors between the point clouds is obviously reduced, for example, the mean value of the registration errors between the point clouds 4 and 12 is reduced from 43 meters to 5.2 meters, and the standard deviation is also reduced by different magnitudes, which indicates that the registration errors are more concentrated.

TABLE 3 Global GPA+ICP registration front and rear Point cloud bias statistics table

And secondly, evaluating the absolute accuracy of the point cloud set before and after registration with the reference point cloud, and evaluating the absolute registration accuracy by using the LOLA laser checking points. The total of 710 laser checking points are selected, absolute deviation related statistics between point clouds before and after registration and the LOLA laser checking points are shown in table 4, the absolute deviation RMSE is reduced from 30.13m to 3.86m, and meanwhile, the standard deviation is also greatly reduced, so that the overall deviation is greatly reduced and is more concentrated. For isolated point cloud registration, absolute deviation related statistics between isolated point clouds and the LOLA laser verification points before and after registration can be obtained as shown in table 5 below: the error between the Shan Jingdian cloud before registration and the reference point cloud is different, for example, the error of the point cloud 2 is smaller, and the errors of the point clouds 3 and 5 are larger, and the situation is mainly influenced by the accuracy of the outside; after registration, the absolute deviation mean value of the single-view point cloud and the LOLA laser point is about 3 meters, the absolute deviation is 3-4 meters, the errors are obviously reduced, and the registration effect of the single-view point cloud is good and the single-view point cloud is aligned with the reference point cloud. All the point clouds are globally regulated through the global registration strategy, so that the self-consistency and absolute precision of the spatial positions of the point clouds are ensured.

TABLE 4 Table of absolute deviation between point clouds DEM and LOLA laser verification points before and after registration

TABLE 5 Absolute bias table between the foreground DEM and LOLA laser check points before and after registration

And S7, aiming at the multi-view cloud subjected to global registration, obtaining a fusion point cloud by adopting a spatial grid fusion strategy of mixed weighting of a triangulation error and a distance, and obtaining a fusion DEM by interpolation. The method comprises the following substeps.

S71, obtaining the side length l of the minimum bounding box of the point cloud according to the maximum and minimum values of X, Y, Z coordinate axes of the three-dimensional point set of the multi-view cloud _x 、l _y 、l _z 。；

S72, setting a voxel small grid side length cell, and dividing a minimum bounding box into M x N x L space grids;

s73, numbering each space grid, and determining the space grid to which each three-dimensional point belongs;

s74, calculating a space grid fusion value by adopting a fusion strategy considering the mixed weighting of the triangulation error and the distance.

The weight formula and the grid value calculation formula of the space grid value are as follows:

wherein w is _i,t Represents the ith ^th Point-to-t ^th Weights in grid, d _i,t And e _i,t The euclidean distance and triangulation error between the i-th point and the t-th grid center are represented, respectively. Alpha (alpha) ₁ Sum alpha ₂ The weight values of the relative error and distance terms are triangulated, respectively. z _t To fuse the spatial grid values.

The weight formula of the invention is respectively provided with alpha ₁ ＝0.5，ɑ ₂ =0.5. And obtaining a fused terrain model through fusion and interpolation.

S8, slope-based outlier detection and contour detection-based hole filling are carried out on the fusion DEM.

Specifically, the abnormal grid detection formula based on the gradient is as follows:

Thresholdx＝k·RMSEx _DSC

Thresholdy＝k·RMSEy _DSC

wherein DSC (Differences in Slope Change) is a single point gradient difference, and the RMSE value of the sum of each grid is calculated by calculating the sum of DSC values of each data point in the same direction _DSC . The threshold is set to k times the RMSE in each direction.

First, on the abnormality detection side, taking k=3, threshold=0.196135, threshold= 0.197602. 865 pixels are retrieved as coarse differences in 18605325 grids of the fused DEM total. Fig. 6 is an abnormality detection case of the enlarged area.

Secondly, in the aspect of hole filling, according to a binary image fused with the DEM, all contours are detected and stored in a linked list, contours are output in a Freeman link code mode, and linear interpolation filling is carried out on the two directions of x and y. Fig. 7 is a schematic diagram of hole filling in an enlarged area. Through the post-processing steps, the DEM and the mountain shadow map are finally fused as shown in fig. 8.

Example 2

On the basis of embodiment 1, this embodiment provides an electronic device, including: the system comprises one or more processors and a memory, wherein one or more programs are stored in the memory, and the one or more programs comprise instructions for executing the global registration method of the moon point cloud based on the overlapping degree classification.

The method and the device adopt a global registration method to adjust the multi-view point cloud under the condition that prior knowledge of view sequence is not needed, and meanwhile, aiming at the complex distribution condition of the multi-view point cloud, classification is carried out according to the overlapping degree, and the global registration method combining global GPA+ICP adjustment and ICP registration based on the reference point cloud is adopted. On the point cloud fusion side, a fusion strategy of considering the mixed weighting of the triangulation error and the distance is adopted, so that the fusion of the multi-view point clouds is realized, the fine contradiction among the point clouds is effectively eliminated, the data redundancy is removed, and meanwhile, the integrity and the precision of the fusion terrain model can be ensured by carrying out post-processing on the fusion terrain model.

While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.

Claims (10)

1. The global registration method of the moon surface point cloud based on the overlap degree classification is characterized by comprising the following steps of:

acquiring a plurality of overlapped or partially overlapped point clouds, extracting the boundary of each point cloud, calculating the overlapping degree between the point clouds, and dividing the point clouds into an overlapped point cloud set and an isolated point cloud set based on the overlapping degree and a preset threshold value;

acquiring a reference digital terrain model of a moon surface, obtaining reference point clouds through dispersion, and carrying out density consistency processing on each single-view point cloud in the overlapped point cloud set and the isolated point cloud set;

performing iterative registration with the reference point cloud based on global generalized general analysis and iterative nearest point algorithm aiming at the overlapped point cloud set;

registering the isolated point cloud set with the reference point cloud based on an iterative nearest point algorithm;

and (3) completing global point cloud registration to obtain the multi-view cloud after global registration.

2. The global registration method of the lunar surface point cloud based on the overlapping degree classification as claimed in claim 1, wherein the calculating process of the overlapping degree between the point clouds comprises the following steps:

and projecting the point cloud on a two-dimensional plane, acquiring a boundary point set of the point cloud based on an alhpa shapes algorithm, simplifying the boundary point set based on a convex hull algorithm, and calculating the overlapping degree between the point clouds through a geos library.

3. The global registration method of lunar surface point cloud based on overlap classification of claim 1, wherein the density matching process comprises:

and enabling the density of each single-view point cloud in the overlapped point cloud set and the isolated point cloud set to be consistent with the density of the reference point cloud through voxel filtering.

4. The global registration method of lunar surface point cloud based on overlap classification according to claim 1, wherein the process of registering the overlapping point clouds comprises:

searching the nearest matching relation in the rest point clouds aiming at each point in each point cloud, calculating a rotation matrix and a translation matrix of each point cloud based on the nearest matching relation and a preset objective function, updating the point clouds, and repeatedly executing the steps until a preset global convergence condition is met, thereby completing registration.

5. The global registration method of the lunar surface point cloud based on overlap classification as claimed in claim 4, wherein the global convergence condition is:

the iteration times exceed a preset value, or the global registration errors before and after global registration are smaller than the preset value, wherein the global registration errors are based on the average value of the corresponding point distances from the point cloud to the centroid cloud, and the centroid cloud is calculated based on the independent subset of each nearest matching relation.

6. The global registration method of lunar surface point cloud based on overlap classification as claimed in claim 4, wherein said objective function is:

wherein m is the number of overlapping point clouds, A ₁ ,A ₂ ,…,A _m Respectively representing m point sets, p is the number of point clouds in the point cloud set, and letT is a rotation matrix, j is a p×1 unit vector, T is a translation matrix, and p×p matrix M _i Representing the visibility of the point set, 1 being visible, 0 being absent, < ->

7. The global registration method of lunar surface point clouds based on overlap classification of claim 1, further comprising:

based on the multi-view cloud after global registration, a spatial grid fusion strategy based on mixed weighting of triangulation errors and distances is adopted to obtain fusion point cloud, and interpolation is carried out to obtain a fusion digital terrain model;

and carrying out gradient-based outlier detection and contour detection-based hole filling on the fused digital terrain model to finish fusion of the point cloud and the digital terrain model.

8. The global registration method of lunar surface point cloud based on overlap classification of claim 7, wherein the process of obtaining the fused digital terrain model comprises the following steps:

calculating the size parameter of a minimum bounding box of the point cloud based on extreme values on three coordinate axes of the multi-view cloud three-point set after global registration;

dividing the minimum bounding box into a plurality of space grids based on the set voxel small grid side length and numbering;

and calculating a space grid fusion value by adopting a fusion strategy of mixed weighting of the triangulation error and the distance to obtain the fused digital terrain model.

9. The global registration method of lunar surface point clouds based on overlap classification as claimed in claim 8, wherein the calculation of the spatial grid fusion value is implemented by the following formula:

wherein w is _i,t Represents the ith ^th Point-to-t ^th The weight in the grid is determined by the weight in the grid,representing the spatial grid fusion value, d _i,t And e _i,t Respectively representing Euclidean distance and triangulation error between ith point and tth grid center, alpha ₁ Sum alpha ₂ The weight values of the relative error and distance terms are triangulated, respectively.

10. An electronic device, comprising: one or more processors and memory, the memory having stored therein one or more programs, the one or more programs comprising instructions for performing the overlapping degree classification-based moon point cloud global registration method of any of claims 1-9.