CN114240727B

CN114240727B - Polygon topology generation method and device based on GPU acceleration

Info

Publication number: CN114240727B
Application number: CN202111446440.6A
Authority: CN
Inventors: 杨科; 朱泳标; 张广泽; 李娜; 邹杨; 吴彦格; 陈兵; 黄弘; 肖红玉
Original assignee: China Railway Eryuan Engineering Group Co Ltd CREEC
Current assignee: China Railway Eryuan Engineering Group Co Ltd CREEC
Priority date: 2021-11-30
Filing date: 2021-11-30
Publication date: 2023-06-20
Anticipated expiration: 2041-11-30
Also published as: CN114240727A

Abstract

The invention relates to a polygon topology generation method and equipment based on GPU acceleration, comprising the following steps: extracting a plurality of point data and arc segment data from the layer data, generating a first point set and a first arc segment set, and transmitting the first point set and the first arc segment set to the GPU; generating cluster data according to the first point set and the first arc segment set by adopting a parallel cluster algorithm in the GPU; replacing the first point set according to the cluster data returned by the GPU, further generating a second point set, and generating a second arc segment set according to the second point set; transmitting the second point set and the second arc segment set to the GPU; generating a polygon set according to the second point set and the second arc segment set by adopting a GPU parallel polygon topology generation method in the GPU; and performing de-duplication processing on the polygon set to obtain the polygon set corresponding to the layer data. According to the method, the polygon topology corresponding to the layer data is automatically generated through CPU+GPU parallel calculation, so that human intervention is reduced, and the processing efficiency is improved.

Description

Polygon topology generation method and device based on GPU acceleration

Technical Field

The invention relates to the field of computer graphics layer data processing, in particular to a polygon topology generation method and equipment based on GPU acceleration.

Background

The generation of arc segment-polygon topologies in the layer data has important research and application significance in geology and engineering geology. With the popularization of CAD, plane diagrams, longitudinal section diagrams and cross section diagrams in engineering projects are completed under the assistance of a computer, the dot line and plane construction of the geologic bodies is completed, and the spatial relationship among the geologic bodies is found by means of an arc segment-polygon generation algorithm, so that mapping staff and engineering geology staff are helped to judge landform units, know the spatial distribution conditions of stratum and fault, and provide basic technical support for future intelligent analysis. The research of a high-reliability, high-robustness and high-performance arc segment-polygon generation method is likely to become a key ring in the geographic geological informatization process.

The geographic geological informatization starts earlier, taking domestic large-scale design institute as an example, the current design drawing mainly comprising CAD files reaches the TB level, and the minimum single CAD file also comprises MB as a measurement unit, so that very high requirements are put forward on the performance of the graph layer data arc segment-polygon generation method. Because of the dependency relationship between the points and the arc segments, the existing topology generation method is based on a serial algorithm, judges the dependency relationship between the points and the arc segments to determine the direction of polygon surrounding, and adds the arc segments or the points into the arc segment set of the polygon one by one. The serial mode is not very difficult to process for a plurality of polygons composed of tens of arc segments, but if the serial mode is applied to process tens of arc segments in a conventional CAD drawing, the conditions of time consumption in calculation and exceeding operation load exist, and even the conditions of insufficient memory and the like can be caused, so that the use of a user is influenced.

The GPU acceleration technology mainly utilizes more physical processing cores on a display card and faster access speed, replaces the program which is designed in series in the past with the program which is designed in parallel, thereby improving the overall processing performance, and particularly provides a computing platform for the application development of GPU acceleration and parallel computing by taking CUDA toolkit of NVIDIA company as the main component. Practice proves that the speed of serial processing can be effectively improved by adopting GPU acceleration and parallel computing.

Disclosure of Invention

The invention aims to solve the problems that the existing topology generation method adopts a serial algorithm and is difficult to process a large amount of arc segment data, and provides a polygon topology generation method and equipment based on GPU acceleration.

In order to achieve the above object, the present invention provides the following technical solutions:

a polygon topology generation method based on GPU acceleration, the method comprising:

step 1: extracting a plurality of point data and arc segment data from the layer data, generating a first point set and a first arc segment set, and transmitting the first point set and the first arc segment set to the GPU; in the GPU, a parallel clustering algorithm is adopted to generate a cluster and a centroid thereof according to the first point set and the first arc segment set;

step 2: replacing the first point set according to the cluster returned by the GPU and the centroid thereof, further generating a second point set, and generating a second arc segment set according to the second point set; transmitting the second point set and the second arc segment set to the GPU; generating a polygon set according to the second point set and the second arc segment set by adopting a parallel polygon topology generation method in the GPU;

step 3: and performing de-duplication processing on the polygon set to obtain a polygon set corresponding to the layer data.

According to the polygon topology generation method based on GPU acceleration, the bottleneck problem in the polygon generation process is analyzed, the existing calculation flow is modified, and an algorithm which is more beneficial to parallel calculation is selected, so that the performance problem of the whole arc segment-polygon generation can be greatly improved according to the Almdar law.

According to a specific embodiment, in the polygon topology generation method based on GPU acceleration, each arc segment in the arc segment set is arranged according to the sequence number of the points forming the arc segment in the point set; each polygon in the polygon set is arranged according to the sequence numbers of the points constituting the polygon in the point set.

According to a specific embodiment, in the method for generating a polygon topology based on GPU acceleration, in the GPU, a parallel clustering algorithm is adopted to generate a cluster and a centroid thereof according to the first point set and the first arc segment set, including:

traversing the first arc segment set to obtain line segments among points on the arc segments to obtain line segment information; calculating line segment intersection points by using a plurality of parallel first lines according to the line segment information and the first point set, and generating an intersection point set; combining the intersection point set and the first point set to obtain a third point set;

and clustering the third point set by adopting a parallel QT clustering algorithm to obtain the cluster and the mass center thereof.

According to a specific embodiment, in the method for generating a polygon topology based on GPU acceleration, the method further includes:

and reserving a storage space for the intersection point set in a global storage area of the GPU according to the line segment information before parallel intersection calculation is carried out on the line segment information and the first point set.

According to a specific embodiment, in the method for generating a polygon topology based on GPU acceleration, the clustering processing is performed on the third point set by using a parallel QT clustering algorithm to obtain the cluster and a centroid thereof, including:

calculating Euclidean distances between each point in the third point set and other points by using a plurality of parallel second threads, sequentially judging threshold values of Euclidean distances obtained by the plurality of second threads by using the third threads, respectively obtaining relevant points, corresponding to each point, of which the distance is smaller than the threshold value, and selecting the point with the most relevant points as a clustering point; the selected cluster points and the related points form a cluster, and the coordinates of the cluster points and the related points are utilized to solve the barycenter coordinates.

According to a specific embodiment, in the method for generating a polygon topology based on GPU acceleration, the replacing the first point set according to the cluster returned by the GPU and the centroid thereof, further generating a second point set, includes:

and replacing the coordinates of the cluster points and the related points in the first point set by using the coordinates of the centroid, so as to generate the second point set.

According to a specific embodiment, in the method for generating a polygon topology based on GPU acceleration, in the GPU, a parallel polygon topology generating method is adopted to generate a polygon set according to the second point set and the second arc segment set, including:

generating an adjacency matrix according to the line segment information in the second point set and the second arc segment set;

traversing each point in the adjacency matrix by using a plurality of parallel fourth threads, and sequentially finding out an adjacency relation corresponding to each point, wherein the adjacency polygons generated by each fourth thread form the polygon set.

In a further embodiment of the present invention, there is also provided an electronic apparatus including:

one or more processors;

and a memory for storing one or more programs that, when executed by the one or more processors, cause the one or more processors to implement the GPU-accelerated polygon topology generation method described above.

Compared with the prior art, the invention has the beneficial effects that:

according to the polygon topology generation method based on GPU acceleration, a CPU and GPU cooperative processing mode is adopted, and aiming at the data flow characteristics of each step of the polygon topology data generation flow, the data processing of each step is distributed in different layers (CPU and GPU) to be automatically executed, so that the performance problem of the traditional arc segment polygon generation is effectively improved; meanwhile, corresponding parallel threads are configured in the GPU, so that the data processing efficiency is improved, and the geographic geological topology generation efficiency and the geographic geological topology generation automation degree are practically improved.

Drawings

FIG. 1 is a flowchart of an arc segment polygon generation method based on GPU parallel computing according to an embodiment of the invention;

FIG. 2 is a flowchart of a GPU parallel computing intersection point according to an embodiment of the present invention;

FIG. 3 is a flowchart of QT cluster parallel computation in a GPU according to an embodiment of the present invention;

fig. 4 is a flowchart of parallel computing polygon topology in a GPU according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to test examples and specific embodiments. It should not be construed that the scope of the above subject matter of the present invention is limited to the following embodiments, and all techniques realized based on the present invention are within the scope of the present invention.

Example 1

FIG. 1 shows a GPU parallel computing based arc segment polygon generation method according to an exemplary embodiment of the present invention, comprising the steps of: parallel computing intersection points, parallel QT clustering, parallel searching and generating polygons:

1. the original point set and the original arc segment set data are extracted from the CAD file (wherein the arc segment set comprises a plurality of pieces of line segment information forming each arc segment). In practical application, mapping and geology personnel conduct surveying on site, CAD files of plane, cross section, vertical section and topography map are generated by utilizing the existing algorithm, in mapping and geology profession, different terrains and strata are distinguished in a layer mode. The set of points and the set of arcs in the host memory (CPU) are copied into the device memory (GPU). Host memory and device memory are terms defined in CUDA development to distinguish between memory accessed by a CPU and memory accessed by a GPU in a computer.

2. Space is allocated in advance for the intersection points among all the line segments to be generated in parallel from a global storage area in the equipment memory, namely, the line segments (the serial number information of the two end points of the line segments and the coordinate information of the points) among the points of the arc segments are obtained by traversing the points forming the arc segments, and the combination number is calculated for all the line segments

(where n=the number of line segments), the reserved intersection space is allocated according to the number of combinations.

3. The intersection points are calculated in parallel in the GPU, as shown with reference to fig. 2. According to the point set and the line segment information, utilizing the coordinate information of the end points forming the line segments to solve the intersection points among the line segments, and for the intersection points which do not fall in any line segment, if the distance between the intersection point and a certain line segment is within a threshold value range, the intersection point set is included, otherwise, the intersection point is not included; and correspondingly adding the two data into the reserved intersection point space. And according to the point set and the line segment set, obtaining an intersection point set and the relation between the intersection point and the line segment in the GPU in parallel, and for the intersection point which does not fall in any line segment, if the distance between the intersection point and a certain line segment is within a threshold range, the intersection point set is included, otherwise, the intersection point is not included. And traversing the intersection points in the reserved space, and forming a new point set by the intersection point set and the original point set. The new set of points is then QT clustered in the GPU. QT clustering is an algorithm, in which euclidean distances between a certain point and other points are calculated through a preset threshold, points (related points) with distances smaller than the threshold are counted, and finally the point with the largest count and the related points thereof are adopted as a cluster. And (3) aggregating the points with the largest quantity in the threshold range, and suppressing the minimum value to obtain the cluster center (centroid) of the cluster as the candidate point of the cluster. According to the sequence numbers of points in the clustering result, coordinates of each point are obtained in a point set, and the mass center of the cluster is calculated, wherein the mass center formula is as follows:

wherein i is any point in a cluster, X _i Is the abscissa of the point i, Y _i Is the ordinate of point i. The method carries out parallel optimization on a large number of calculations in QT clusters, and carries out parallel processing on distance calculation in reference to FIG. 3, and for the logic control of quantity judgment, the logic control is processed in a GPU single thread, so that intermediate data of the step is reduced and copied to a host memory, and the purpose of saving processing time is achieved.

4. And (3) utilizing a clustering result (a clustering cluster and a centroid thereof) generated by the GPU to modify the relation between the original point set and the arc segment set, replacing all points in the clustering by using the clustering centroid, and updating the clustering centroid to the point set and the arc segment set to form new point set and arc segment set data. The step needs a large amount of logic judgment, is not suitable for calculation in the GPU, copies the clustering result from the equipment memory to the host memory, and is convenient for updating the point set and the arc segment set by using the CPU later. Traversing the point set according to the sequence numbers of the points in the clustering result, replacing the coordinates in the point set with the coordinates of the mass center according to the sequence numbers of the points in the clustering result, and replacing the sequence numbers of the points in the arc segment set with the sequence numbers of the mass center in the point set. The loop is continued until all points in the cluster have been processed. Traversing the point set, reducing the same points, and generating a new point set after de-duplication. Traversing the arc segment set, and taking the sequence number in the new point set as the point sequence number in the new arc segment set. Since this process involves a large number of logical branching operations, placing the calculation in the host computer can properly improve the efficiency of the program. Through the process, the number of points in the original point set is greatly reduced, and then the new point set and the new arc segment set are copied to the equipment memory.

5. And generating an adjacency matrix according to the connection relation between the point set and the arc segment concentrated points. And the space is pre-allocated for the polygons from the global storage area in the equipment memory, the polygons in the drawing are arc segments divided by 7 according to the experience data, and the value can be dynamically adjusted according to the complexity of the drawing. Referring to fig. 4, this step mainly uses multiple parallel single threads to traverse the adjacency matrix with any point in the adjacency matrix as a starting point, and finds a non-zero adjacency relation that each starting point is connected in turn: the traversal end condition is that the end point is the start point (constituting the polygon area), or that all adjacency traversal ends (not constituting the polygon area). The polygon generated by each thread is an array of sequence numbers of points of the polygon.

Specifically, by line segment information, an adjacency relation is marked in an adjacency matrix (M), for example: if point P1 is connected to point P2, [ P1, P2] and [ P2, P1] in the matrix are 1, otherwise 0, P1 and P2 can be the same point. The line segment connection information is not much, so that the adjacent matrix is a sparse symmetric matrix; in the GPU, a single computing unit (SM) circulates adjacent relations which are not 0 in the upper triangle part of the sparse adjacent matrix, all adjacent relations which are not 0 in the end point are taken as the current adjacent relations of the next circulation according to the end point of the current adjacent relations, and the points are added into the point set of the area. The cycle end condition is that the end point is the start point (constituting the polygon area), or that all adjacency traversals end. For example: and (3) taking P1 as a starting point to circulate, finding a non-zero adjacent relation corresponding to P1, assuming that the found points are P7, P8 and P16, taking P7, P8 and P16 as the starting points to find the adjacent relation of P7, P8 and P16 in the next circulation traversal, and forming a polygon until the end point of the found adjacent relation is P1, or finishing traversal of all points (without returning to the polygon area formed by P1).

And taking all adjacencies which are not 0 as the current adjacencies of the next cycle according to the end point of the current adjacencies, and adding the point into the point set of the area. The cycle end condition is that the end point is the start point (constituting the polygon area), or that all adjacency traversals end.

6. Since the previous steps are all independent and parallel calculations, a large number of repeated polygons are generated, the generated polygons need to be de-duplicated, the steps involve relatively complex logic operations, and the CPU needs to be used for judgment and calculation in the host. Copying the polygon result to the host memory. In the CPU, the polygons are traversed, and the latter need to be deduplicated for the same ring sequences, such as {1,2,3,4,5} and {2,3,4,5,1}, which can be considered the same polygon. The polygon set after the duplication removal is the final result. And calling the components developed based on Object-Arx according to the final polygons, and marking the polygons on the CAD drawing by different colors, so that mapping and engineering geology personnel can be assisted to view and identify.

Example 2

In a further embodiment of the present invention, an electronic device (e.g., a computer server with a program execution function) for implementing the polygon topology generation method based on GPU acceleration is provided, which includes at least one processor, a power supply, and a memory and an input/output interface communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method disclosed in any one of the preceding embodiments; the input/output interface can comprise a display, a keyboard, a mouse and a USB interface, and is used for inputting and outputting data; the power supply is used for providing power for the electronic device.

Those skilled in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.

The above-described integrated units of the invention, when implemented in the form of software functional units and sold or used as stand-alone products, may also be stored in a computer-readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in essence or a part contributing to the prior art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims

1. A polygon topology generation method based on GPU acceleration, the method comprising:

2. The GPU acceleration-based polygon topology generation method of claim 1, wherein each arc in the set of arcs is arranged according to a sequence number of points in the set of points that make up the arc; each polygon in the polygon set is arranged according to the sequence numbers of the points constituting the polygon in the point set.

3. The GPU acceleration-based polygon topology generation method of claim 1, wherein generating clusters and centroids thereof in the GPU from the first set of points and the first set of arcs using a parallel clustering algorithm comprises:

4. A GPU acceleration based polygon topology generation method as recited in claim 3, further comprising:

5. The GPU acceleration-based polygon topology generation method of claim 3, wherein clustering the third set of points using a parallel QT clustering algorithm to obtain the cluster and a centroid thereof comprises:

calculating Euclidean distances between each point in the third point set and other points by using a plurality of parallel second threads, sequentially judging threshold values of Euclidean distances obtained by the plurality of second threads by using the third threads, respectively obtaining relevant points, corresponding to each point, of which the distance is smaller than the threshold value, and selecting the point with the most relevant points as a clustering point; the selected cluster points and the related points form a cluster, and the mass center of the cluster is solved according to the coordinates of the cluster points and the related points.

6. The GPU acceleration-based polygon topology generation method of claim 5, wherein replacing the first set of points according to the cluster returned by the GPU and the centroid thereof to generate a second set of points comprises:

and replacing coordinates of the cluster points and related points in the first point set by using the barycenter coordinates, so as to generate the second point set.

7. The GPU acceleration-based polygon topology generation method of any of claims 1-6, wherein generating, in the GPU, a polygon set from the second point set and the second arc segment set using a parallel polygon topology generation method comprises:

8. An electronic device, the electronic device comprising:

one or more processors;

a memory for storing one or more programs that, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-7.