WO2021232463A1

WO2021232463A1 - Multi-source mobile measurement point cloud data air-ground integrated fusion method and storage medium

Info

Publication number: WO2021232463A1
Application number: PCT/CN2020/092817
Authority: WO
Inventors: 陈琳海
Original assignee: 北京数字绿土科技有限公司
Priority date: 2020-05-19
Filing date: 2020-05-28
Publication date: 2021-11-25
Also published as: CN111598823B; CN111598823A

Abstract

A multi-source mobile measurement point cloud data air-ground integrated fusion method and a storage medium. The fusion method comprises: extracting building contour point cloud data A from point cloud data collected by different platforms in a clustering judgment mode, and extracting, by means of two-dimensional analysis of three-dimensional gridding, rod-shaped object point cloud data B from point cloud data collected by different platforms; matching features of the building contour point cloud data A or the rod-shaped object point cloud data B extracted from different platforms, and determining whether the features relate to the same building contour or the same rod-shaped object or not; and registering, by means of adjustment calculation according to the feature matching result, the point cloud data collected by different platforms to obtain the registration fusion result. According to the fusion method, point cloud registration of different platforms is carried out by adopting object-level features, and the applicability is higher, the precision is higher, and point cloud data collected by multiple platforms is subjected to fusion management, and integration of multi-time-space and multi-platform data is realized.

Description

Multi-source mobile measurement point cloud data air-ground integrated fusion method and storage medium

Technical field

The invention relates to the technical field of inspection and surveying, in particular to a method and storage medium for integrating multi-source mobile measurement point cloud data and air-ground integration.

Background technique

Lidar is a technology for quickly obtaining three-dimensional point cloud data on the surface of an object. It has become a major technical means for three-dimensional earth observation with high temporal and spatial resolution. An increasingly important role. In terms of data acquisition platforms, it includes satellite platforms, airborne platforms, vehicle-mounted platforms, ground/piggyback platforms, and so on.

At present, most technologies for obtaining 3D point cloud data through lidar use a single platform for point cloud data acquisition. Due to the limited observation range of a single viewing angle and a single platform and the inconsistent spatial reference, in order to obtain all-round spatial information of the target area, it is not only necessary For point cloud data fusion between stations/strips, point cloud data fusion on multiple platforms (such as airborne, vehicle-mounted, ground, etc.) is also required to make up for the lack of data caused by a single perspective and a single platform to achieve a large range A complete and detailed description of the digital reality of the scene; the integration of point cloud data from different platforms requires the association of features with the same name. Currently, the common features of existing methods are low-level features such as normal vectors, key points, and point feature histograms. However, due to scanning on different platforms Different viewing angles, different coverage areas, and huge differences in the point density of the scanned point cloud data. The extracted low-order features are greatly affected by the point distribution and point density, and the algorithm accuracy is not high and the robustness is poor.

The Chinese patent with the application number CN201410047608.X published the invention patent of "a multi-platform point cloud data fusion method". The main point is to analyze the accuracy of the point cloud data after filtering and denoising, based on the point cloud data with the highest accuracy. , Perform accuracy correction on the remaining data. The specific implementation method is to compare the accuracy of the preprocessed point cloud data, and based on the point cloud data with higher accuracy, correct and analyze the lower accuracy data to obtain the point cloud data. Convert the model and perform correction and fusion. The specific processing steps are: 1) Determine the change of point cloud data in the same area by constructing a digital surface model according to the date of point cloud data collection; 2) Extract and correct the point cloud data change range after detection Point, generate an updated model; 3) update the point cloud data with poor accuracy according to the updated model; 4) check the accuracy of data fusion by constructing a digital surface model); this technical solution is mainly based on the change of point cloud data at different times Update the data model model, and then update the point cloud data with poor accuracy according to the updated model. Although the point cloud data with poor accuracy can be updated, it will also cause a certain degree of distortion of the overall model, and it is consistent with the present invention. There are essential differences in the methods based on the special integration of multiple object levels.

The Chinese patent with the application number CN201910266150.X published the invention patent application of "Multi-platform point cloud intelligent processing method for holographic surveying and mapping". The main point is the high-precision fusion of laser point cloud data from multiple platforms. The specifics are: nearest neighbor point cloud search , The construction of the global matching energy equation and the minimum cost matching method of the bipartite graph, but the overall method is more complicated, the calculation amount is large, and it is fundamentally different from the method of the present invention based on the special fusion of multiple object levels. Therefore, how to solve the current defects of fusion registration based on 3D point cloud data obtained from different platforms is a problem that needs to be solved at this stage.

Summary of the invention

The purpose of the present invention is to overcome the shortcomings of the prior art and provide a multi-source mobile measurement point cloud data air-ground integrated fusion method, storage medium and terminal, which solves the current problem of fusion and registration based on three-dimensional point cloud data acquired on different platforms. Existing shortcomings, as well as the current shortcomings of using low-level features for registration.

The object of the present invention is achieved by the following technical solutions: a method for integrating air-ground integration of multi-source mobile measurement point cloud data, and the fusion method includes:

Extract the building contour point cloud data A from the point cloud data collected on different platforms through clustering judgment, and extract the rod-shaped points from the point cloud data collected on different platforms through the three-dimensional grid-based two-dimensional analysis Cloud data B;

Match the features of building contour point cloud data A or rod point cloud data B extracted from different platforms to determine whether it is the same building contour or the same rod;

According to the feature matching result, the point cloud data collected by different platforms are registered through adjustment calculation, and the registration fusion result is obtained.

The fusion method further includes the step of collecting point cloud data in the same area through multiple different platforms before extracting the building contour point cloud data A and the rod-shaped point cloud data B.

The step of extracting building profile point cloud data A from the point cloud data collected on different platforms by means of clustering judgment includes the following:

Project the point cloud into the three-dimensional coordinate system to obtain a projection base surface, and the projection base surface is perpendicular to any coordinate plane;

The projection points on the projection base surface are divided into different projection surface point cloud block sets according to a certain interval;

Calculate the height difference of the adjacent projection surface point cloud block sets according to the highest point of each projection surface point cloud block set, and judge whether the two projection surface point cloud block sets belong to the same cluster according to the calculation result;

The random sampling consensus algorithm is used to fit the projection point cloud of each cluster, extract the building contour line segment from the projection point cloud of each cluster, and record the start and end coordinates of the building contour line segment.

The highest point of the certain projection surface point cloud block set is the projection point cloud with the farthest distance in the coordinate axis direction parallel to the projection base plane; if the height difference of the adjacent projection surface point cloud block sets is less than the standard value, then It shows that the two projection surface point cloud block sets are continuous features and belong to a cluster.

The step of extracting rod point cloud data B from point cloud data collected on different platforms through three-dimensional grid-based two-dimensional analysis includes the following:

The point cloud data collected by different platforms are three-dimensionally gridded, and the grid is segmented and clustered through connectivity analysis. According to the clustering area with an area smaller than the threshold value and the screening of the cross-sectional shape and the length of the main axis, the potential rod shape is obtained Object category and non-rod category;

Calculate the centroid coordinates of each cluster in the potential rod category, and set an inner radius and an outer radius with the centroid coordinates as the point. The inner radius contains all the point clouds of the cluster, the inner radius and There is no point cloud in the circle formed by the outer radius;

Set the minimum height threshold of a shaft. If the height data of a certain cluster of potential shaft types is greater than the minimum height threshold, it is judged as a shaft and the starting and ending coordinates of the shaft are recorded. , To extract the line segment of the rod.

The screening of the cross-sectional shape and the length of the main axis includes: judging whether the cross-sectional shape is circular, and if it is, it is described as a rod; judging whether the length of the main axis is less than a specified value, and if it is, it is described as a rod.

The matching the features of the building contour point cloud data A or the rod point cloud data B extracted from different platforms, and determining whether it is the same building contour or the same rod includes:

The features extracted from different platforms are respectively denoted as LA={LA _i ,i=1, 2,...,m} and LB={LB _i ,i=1, 2,...,n}, and from LA and LB respectively Select two building outlines or rods and mark them as LA ₁ , LA ₂ , LB ₁ and LB ₂ ;

Calculate the angle, distance and length difference of two building outlines or poles;

And judge whether the calculation result meets the conditions. If the conditions are met, it means that the two building outlines or poles are the same building or poles in different platforms.

The registration and fusion result of the point cloud data collected by different platforms through adjustment calculation according to the feature matching result includes:

Extracting the three-dimensional coordinate information of the two building outlines or poles that are determined to be the same building or pole to calculate the rotation parameter, the translation parameter, and the zoom parameter;

Take the three-dimensional coordinates of the point cloud collected by any platform as the target coordinate system, and rotate, translate and scale the original coordinate system of the point cloud collected by another platform to make the three-dimensional coordinates of the point cloud in the original coordinate system and the target coordinate system The three-dimensional coordinates of the midpoint cloud coincide to realize the registration and fusion of point cloud data from different platforms.

A storage medium, characterized in that: a computer program is stored in the storage medium, and when the computer program is running, the steps of a method for integrating air and ground of multi-source mobile measurement point cloud data are executed.

A terminal includes a memory, a processor, and a control program based on the integrated air-ground integration of multi-source mobile measurement point cloud data that is stored on the memory and can run on the processor, and the multi-source mobile measurement point is based on The control program of cloud data and air-ground integration executes the steps of a multi-source mobile measurement point cloud data and air-ground integration method when running.

The present invention has the following advantages: a multi-source mobile measurement point cloud data and air-ground integrated fusion method, storage medium and terminal, according to the data collected from different platforms to classify objects such as building contours and rods in the same space-time area The feature extraction and feature matching are performed to determine whether the same building outline or the same rod is judged. Finally, the point cloud data collected by different platforms is registered through the adjustment calculation, and the registration fusion result is obtained, so that the final result is The results are more accurate and can be applied to more occasions; and the integration of multi-platform collection point cloud data management has realized the integration of multi-temporal and multi-platform data.

Description of the drawings

Figure 1 is a schematic flow diagram of the method of the present invention;

FIG. 2 is a schematic diagram of the process of extracting point cloud data A of the building outline according to the present invention;

Fig. 3 is a schematic diagram of the process of extracting rod-shaped point cloud data B according to the present invention.

Detailed ways

In order to make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments These are a part of the embodiments of the present invention, but not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein may be arranged and designed in various different configurations.

Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

The present invention will be further described below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.

As shown in Fig. 1, an integrated air-ground fusion method for multi-source mobile measurement point cloud data, the fusion method includes:

Collect point cloud data in the same space area by mounting lidar scanners on multiple platforms such as drones and vehicles;

Further, as shown in FIG. 2, the step of extracting building contour point cloud data A from point cloud data collected on different platforms by means of clustering judgment includes the following contents:

(1) Project the point cloud into the three-dimensional coordinate system to obtain a projection base surface, and the projection base surface is perpendicular to the XOY coordinate plane;

(2) The projection points on the projection base surface are divided into different projection surface point cloud block sets at intervals of 1.5 times the average point spacing;

(3) Calculate the height difference of the adjacent projection surface point cloud block sets according to the highest point of each projection surface point cloud block set, and judge whether the two projection surface point cloud block sets belong to the same cluster according to the calculation result;

Among them, the highest point of a certain projection surface point cloud block set is the projected point cloud with the YOZ coordinate plane perpendicular to the XOY coordinate plane along the Z axis direction with the farthest distance; if the height difference of the adjacent projection surface point cloud block sets is less than the standard value , It means that the two projection surface point cloud block sets are continuous features and belong to a cluster; otherwise, they are regarded as a new category.

(4) Fit the projection point cloud of each cluster by random sampling consensus algorithm, extract the building contour line segment from the projection point cloud of each cluster, and record the start and end coordinates of the building contour line segment .

Further, as shown in FIG. 3, the step of extracting rod point cloud data B from point cloud data collected on different platforms through a three-dimensional grid-based two-dimensional analysis includes the following:

(1) The point cloud data collected by different platforms is three-dimensionally gridded, and the grid is segmented and clustered through connectivity analysis. According to the clustering area with an area less than the threshold and the screening of the cross-sectional shape and the length of the main axis, the potential Types of rods and non-rods;

(2) Calculate the centroid coordinates of each cluster in the potential rod category, and set an inner radius and an outer radius with the centroid coordinates as the circle point. The inner radius contains all the point clouds of the cluster, There is no point cloud in the circle formed by the inner radius and the outer radius; this step is actually an optimization of the result of the previous step, because if it is a rod, its cross-sectional shape is a circle (that is, it can pass through a center and radius Including all points), if there are other points in the surrounding range, it means that the fitting may be vegetation or other features.

(3) Set the minimum height threshold of a rod. If the height data of a certain cluster of the potential rod type is greater than the minimum height threshold, it is judged as a rod and the starting point of the rod is recorded And the end point coordinates (or the diameter of the shaft can be recorded), and the line segment of the shaft can be extracted. As shown in Figure 3, the step of extracting the rod point cloud data B from the point cloud data collected by different platforms through the two-dimensional analysis of the three-dimensional grid is a complex program execution step (which specifically involves the point cloud Data three-dimensional grid processing, judge and evaluate according to parameters such as cross-sectional shape, main shaft length, etc.), but it is still a subsidiary step of step S2, which is one of the important technical protection points.

Generally speaking, the height of the minimum shaft is set as the minimum height threshold, and only the height of the point cloud data is greater than the minimum height threshold can be judged as a potential shaft.

Further, the screening of the cross-sectional shape and the length of the main axis includes: judging whether the cross-sectional shape is circular, and if it is, it is described as a rod; judging whether the length of the main axis is less than a specified value, and if it is, it is described as a rod. Things.

Further, the matching the features of the building contour point cloud data A or the rod point cloud data B extracted from different platforms, and determining whether it is the same building contour or the same rod includes:

S31. The features extracted from different platforms are respectively recorded as LA={LA _i ,i=1, 2,...,m} and LB={LB _i ,i=1, 2,...,n}, and from LA and LB Select two building outlines or rods respectively in the line and mark them as LA ₁ , LA ₂ , LB ₁ and LB ₂ ;

S32. Calculate the angle, distance and length difference of the two building outlines or poles; this step is to match the same building outline or the same pole in the two platform data, each outline can be As a line segment in the space, judge whether the angle, distance and length difference of each pair of contour lines (that is, each pair of line segments in the space) meet certain conditions. If they are satisfied, the two pairs of contour lines are considered to be the same.

In the above technical solution of the embodiment of the present invention, the feature matching of the same building outline or the same rod in the two platform data can be achieved; for the two building outlines that meet the above matching conditions, the above matching It can be judged that this is based on the same building or the same rod in two different platforms; in the specific technical solution, the above steps of extracting the building outline and extracting the rod will be in the vehicle and airborne point cloud Extract separately, so you will get the contour of the same building and the coordinates of the same rod scanned by different platforms. The result obtained by extracting the building contour is the contour line segment, which records the start and end coordinates of the line segment; the rod extraction result is also a line segment, which records the start and end coordinates, and the diameter of the rod can also be recorded. After feature matching is equivalent to obtaining a series of points with the same name, a mathematical model can be built based on these points with the same name, so feature matching is the technical basis for subsequent establishment of mathematical models.

Further, the registration and fusion result of the point cloud data collected by different platforms through adjustment calculation according to the feature matching result includes:

S41. Extracting three-dimensional coordinate information of the two building outlines or poles that are determined to be the same building or pole to calculate a rotation parameter, a translation parameter, and a zoom parameter;

S42. Take the three-dimensional coordinates of the point cloud collected by any platform as the target coordinate system, and make the three-dimensional coordinates of the point cloud in the original coordinate system and the target coordinate after the original coordinate system of the point cloud collected by another platform is rotated, translated, and scaled. The three-dimensional coordinates of the point cloud in the system coincide to realize the registration and fusion of point cloud data from different platforms.

For example, suppose the coordinates of a point P in the original space are (x, y, z), and its coordinates in the target coordinate system are (X, Y, Z), then rotate it around the z, y, x axis The rotation matrices of γ, β and α angles are:

Thus, the rotation matrix of the coordinate transformation is:

Then the rotated coordinate system is translated x ₀ , y ₀ , z ₀ , so that the origin of the original coordinate system coincides with the origin of the target coordinate system, and finally a scale scaling factor λ is added to complete P from the original coordinate system to the target coordinate system Transformation: In the embodiments of this application, the above-mentioned design algorithm can be used to complete the transformation of P from the original coordinate system to the target coordinate system. The above-mentioned processing procedure is also one of the technical protection points of the embodiments of this application.

Regarding the technical solution of this application, it should be noted that the multi-source mobile measurement point cloud data air-ground integrated fusion method applied in the embodiment of the present invention uses object-level features for point cloud registration of different platforms, which can be based on data from The data collected by different platforms extract and match object-level features such as building contours and rods in the same space-time area to realize whether the same building contour or the same rod is judged, and finally through the adjustment The calculation performs registration on the point cloud data collected by different platforms to obtain the registration fusion result, which makes the final result more accurate and applicable to more occasions.

Point cloud registration on different platforms requires features with the same name. Commonly used features are normal vectors, key points, point feature histograms and other low-level features. The calculation method is to set the search method and neighborhood range for each point. The points in the domain are calculated, and different platforms have different scanning methods, and the obtained point cloud density and point distribution are different. Therefore, the result of registration based on these calculated features is greatly affected by the data situation. However, in the specific technical solution provided by the present invention, the object-level features are used as the features of the same name. The point cloud of these objects can be obtained regardless of whether it is a vehicle or a backpack scan. The point cloud is less affected by the point density and point cloud distribution. Based on these The features are registered, and the result is more reliable. In addition, the air-ground integration method for multi-source mobile measurement point cloud data provided by the embodiment of the present invention can integrate and manage multi-platform collection point cloud data, and realize the integration of multi-temporal and multi-platform data.

Another embodiment of the present invention includes a storage medium, which is characterized in that a computer program is stored in the storage medium, and the computer program executes the steps of a multi-source mobile measurement point cloud data integrated air-ground fusion method when the computer program is running.

Another embodiment of the present invention includes a terminal including a memory, a processor, and a control program that is stored on the memory and can be run on the processor based on the integrated air-ground fusion of multi-source mobile measurement point cloud data, so The control program based on the air-ground integrated fusion of multi-source mobile measurement point cloud data executes the steps of a multi-source mobile measurement point cloud data air-ground integrated fusion method when running.

The present invention uses object-level features as the features of the same name. The point cloud of these objects can be obtained regardless of whether it is vehicle-mounted or backpack scanning, and is less affected by the point density and point cloud distribution. The registration is performed based on these features, and the result is more reliable.

The above are only the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as an exclusion of other embodiments, but can be used in various other combinations, modifications and environments, and It can be modified through the above teaching or technology or knowledge in related fields within the scope of the concept described herein. The modifications and changes made by those skilled in the art do not depart from the spirit and scope of the present invention, and should fall within the protection scope of the appended claims of the present invention.

Claims

An air-ground integrated fusion method for multi-source mobile measurement point cloud data, characterized in that: the fusion method includes:

Extract the building contour point cloud data A from the point cloud data collected on different platforms through clustering judgment, and extract the rod-shaped points from the point cloud data collected on different platforms through the three-dimensional grid-based two-dimensional analysis Cloud data B;

Match the features of building contour point cloud data A or rod point cloud data B extracted from different platforms to determine whether it is the same building contour or the same rod;

According to the feature matching result, the point cloud data collected by different platforms are registered through adjustment calculation, and the registration fusion result is obtained.
The air-ground integrated fusion method of multi-source mobile measurement point cloud data according to claim 1, wherein the fusion method further comprises extracting the building contour point cloud data A and the rod point Before cloud data B, the step of collecting point cloud data in the same area through multiple different platforms.
The air-ground integrated fusion method of multi-source mobile measurement point cloud data according to claim 2, characterized in that: the building contour point cloud is extracted from the point cloud data collected on different platforms by means of clustering judgment The steps of Data A include the following:

Project the point cloud into the three-dimensional coordinate system to obtain a projection base surface, and the projection base surface is perpendicular to any coordinate plane;

The projection points on the projection base surface are divided into different projection surface point cloud block sets according to a certain interval;

Calculate the height difference of the adjacent projection surface point cloud block sets according to the highest point of each projection surface point cloud block set, and judge whether the two projection surface point cloud block sets belong to the same cluster according to the calculation result;

The random sampling consensus algorithm is used to fit the projection point cloud of each cluster, extract the building contour line segment from the projection point cloud of each cluster, and record the start and end coordinates of the building contour line segment.
The air-ground integrated fusion method for multi-source mobile measurement point cloud data according to claim 3, wherein the highest point of the point cloud block set of a certain projection surface is a coordinate axis parallel to the projection base surface The projection point cloud with the farthest direction distance; if the height difference of the adjacent projection surface point cloud block sets is less than the standard value, it means that the two projection surface point cloud block sets are continuous features and belong to a cluster.
A multi-source mobile measurement point cloud data and air-ground integrated fusion method according to claim 4, characterized in that: the rod-shaped data is extracted from the point cloud data collected by different platforms through the two-dimensional analysis of the three-dimensional grid. The steps of point cloud data B include the following:

The point cloud data collected by different platforms is three-dimensional gridded, and the grid is segmented and clustered through connectivity analysis. According to the clustering area whose cross-sectional two-dimensional area is less than the threshold value and the screening of the cross-sectional shape and main axis length, the potential is obtained Types of rods and non-rods;

Calculate the centroid coordinates of each cluster in the potential rod category, and set an inner radius and an outer radius with the centroid coordinates as the point. The inner radius contains all the point clouds of the cluster, the inner radius and There is no point cloud in the circle formed by the outer radius;

Set the minimum height threshold of a shaft. If the height data of a certain cluster of potential shaft types is greater than the minimum height threshold, it is judged as a shaft and the starting and ending coordinates of the shaft are recorded. , To extract the line segment of the rod.
The method for integrated air-ground fusion of multi-source mobile measurement point cloud data according to claim 5, characterized in that: the screening of the cross-sectional shape and the length of the main axis comprises: judging whether the cross-sectional shape is circular, and if so, then The description is a rod; it is determined whether the length of the main shaft is less than the specified value, and if it is, it is described as a rod.
The air-ground integrated fusion method for multi-source mobile measurement point cloud data according to claim 6, characterized in that: the pair of building outline point cloud data A or rod point cloud data B extracted from different platforms Match the features of to determine whether it is the same building outline or the same pole including:

The features extracted from different platforms are respectively denoted as LA={LA i ,i=1, 2,...,m} and LB={LB i ,i=1, 2,...,n}, and from LA and LB respectively Select two building outlines or rods and mark them as LA 1 , LA 2 , LB 1 and LB 2 ;

Calculate the angle, distance and length difference of two building outlines or poles;

And judge whether the calculation result meets the conditions. If the conditions are met, it means that the two building outlines or poles are the same building or poles in different platforms.
The air-ground integrated fusion method of multi-source mobile measurement point cloud data according to claim 7, characterized in that: the point cloud data collected by different platforms are registered according to the feature matching result through adjustment calculation to obtain the registration The results of the fusion include:

Extracting the three-dimensional coordinate information of the two building outlines or poles that are determined to be the same building or pole to calculate the rotation parameter, the translation parameter, and the zoom parameter;

Take the three-dimensional coordinates of the point cloud collected by any platform as the target coordinate system, and make the three-dimensional coordinates of the point cloud in the original coordinate system and the target coordinate after the original coordinate system of the point cloud collected by another platform is rotated, translated, and scaled. The three-dimensional coordinates of the point cloud in the system coincide to realize the registration and fusion of point cloud data from different platforms.
A storage medium, characterized in that: a computer program is stored in the storage medium, and the computer program executes the multi-source mobile measurement point cloud data according to any one of claims 1-8 when the computer program is running. The steps of the chemical fusion method.
A terminal, which is characterized in that it comprises a memory, a processor, and a control program that is stored on the memory and can be run on the processor based on the integrated air-ground integration of multi-source mobile measurement point cloud data. The control program for the air-ground integrated fusion of source mobile measurement point cloud data executes the steps of the method for multi-source mobile measurement point cloud data air-ground fusion according to any one of claims 1-8 when running.