CN106844610B - Distributed structured three-dimensional point cloud image processing method and system - Google Patents
Distributed structured three-dimensional point cloud image processing method and system Download PDFInfo
- Publication number
- CN106844610B CN106844610B CN201710034669.6A CN201710034669A CN106844610B CN 106844610 B CN106844610 B CN 106844610B CN 201710034669 A CN201710034669 A CN 201710034669A CN 106844610 B CN106844610 B CN 106844610B
- Authority
- CN
- China
- Prior art keywords
- point cloud
- dimensional point
- cloud image
- data
- structured
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2228—Indexing structures
- G06F16/2246—Trees, e.g. B+trees
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention provides a distributed structured three-dimensional point cloud image processing method and a system, wherein the method comprises the following steps: acquiring a three-dimensional point cloud image and expansion information corresponding to the three-dimensional point cloud image; classifying the acquired three-dimensional point cloud image and the expansion information thereof, establishing a relational mapping model, and forming a tree-like data structure for storing and nesting a plurality of relational mapping models; storing different relational mapping models as corresponding data files through corresponding processing threads; when a data reading request of the three-dimensional point cloud image is received, establishing a corresponding thread for each stored data file and loading each data file through the corresponding thread to generate corresponding structured data; semantic scene category information of the corresponding scene is obtained according to the structured data, all categories and label codes contained in the scene are analyzed, and the three-dimensional point cloud image is visualized. The invention can efficiently store the three-dimensional point cloud image and the expanded labeling information and efficiently access the structured data in a concurrent manner.
Description
Technical Field
The invention relates to the technical field of images, in particular to the technical field of three-dimensional point cloud images, and specifically relates to a distributed structured three-dimensional point cloud image processing method and system.
Background
The three-dimensional point cloud image is a standardized image data storage data structure in the research and application fields of machine vision, remote sensing modeling, photogrammetry, ancient building reconstruction and the like which need three-dimensional reconstruction and semantic perception labeling of a space scene, the geometric coordinate numerical value of a point on the surface of an object in a three-dimensional space is used as a basic unit for storage, and the whole three-dimensional point cloud data is a set of all the basic units. The three-dimensional point cloud image is used for storing in scene reconstruction and semantic perception labeling tasks at the same time: 1) an original image acquired by a vision sensor; 2) the complete scene image after reconstruction; 3) the image after scene annotation. The point storage basic unit not only contains three Euclidean space coordinate values, but also supports the expansion of color RGB values, normal vectors, curvatures, class labels and the like, and the expansion is directly attached behind the coordinate values and serves as supplementary description information of the current point (as shown in the attached figure-1). Such data structures have significant drawbacks, mainly represented by: 1) repeated storage of extended data for points with homogeneous attributes, with substantial data redundancy, and reduced I/O speed; 2) the random arrangement of the storage units can cause that the whole data needs to be traversed even if a certain category of data in a scene is analyzed independently, and the analysis efficiency is extremely low; 3) marking (manual or automatic) scene information requires a large amount of processing time, and the original three-dimensional point cloud image cannot well fuse the original data of the image and the marking information; 4) for large scenarios, serialized storage schemas are poorly scalable and are limited by the capacity of a single storage device.
A published search of the prior art shows that Martin Weinmann.2016. Reconstructions and Analysis of 3D Scenes From Irregularly Distributed 3D Points to ObjectClasses (1st ed.). Springer Publishing Company, Incorporated. The author discusses how to label the scene information automatically, but no matter the original data or the data with labels, the labels of all the points are stored in the corresponding storage units, and the problems of information redundancy, slow I/O, poor structuralization, poor expansibility and the like cannot be avoided in practical application.
Disclosure of Invention
In view of the above-mentioned shortcomings of the prior art, the present invention aims to provide a distributed structured three-dimensional point cloud image processing method and system for efficiently organizing a structured three-dimensional point cloud image with category label labeling data.
To achieve the above and other related objects, the present invention provides a distributed structured three-dimensional point cloud image processing method, including: acquiring a three-dimensional point cloud image and expansion information corresponding to the three-dimensional point cloud image; classifying the acquired three-dimensional point cloud image and the expansion information thereof, establishing a relational mapping model, and forming a tree-shaped data structure for storing and nesting a plurality of relational mapping models; storing different relational mapping models into corresponding data files through corresponding processing threads for loading and reading; when a data reading request of a three-dimensional point cloud image is received, establishing a corresponding thread for each stored data file and loading each data file through the corresponding thread to generate corresponding structured data; and acquiring semantic scene category information of a scene corresponding to the three-dimensional point cloud image according to the structured data, analyzing all categories and label codes contained in the scene, and visualizing the three-dimensional point cloud image.
In one embodiment of the invention, a JSON key-value storage format is adopted to establish a relational mapping model, and a tree-shaped data structure which is embedded with a plurality of relational mapping models is formed; the data file is a JSON data file.
In an embodiment of the present invention, the path of the data point set of each tag in the tree data structure is described by using a URI.
In an embodiment of the invention, different types of the relational mapping models including the three-dimensional point cloud image and the expansion information thereof are distributively stored in different files and/or different database systems.
In an embodiment of the invention, the extension information includes a color, a category label, a category name, a scene category total number, and a scene name.
In order to achieve the above object, the present invention further provides a distributed structured three-dimensional point cloud image processing system, including: the storage module is used for acquiring the three-dimensional point cloud image and the expansion information corresponding to the three-dimensional point cloud image, classifying the acquired three-dimensional point cloud image and the expansion information thereof, establishing a relational mapping model, forming a tree-shaped data structure for storing and nesting a plurality of relational mapping models, and storing different relational mapping models as corresponding data files for loading and reading through corresponding processing threads; the processing module is used for establishing corresponding threads for each stored data file and loading each data file through the corresponding threads to generate corresponding structured data when receiving a data reading request of the three-dimensional point cloud image, acquiring semantic scene category information of a scene corresponding to the three-dimensional point cloud image according to the structured data, analyzing all categories and label codes contained in the scene, and visualizing the three-dimensional point cloud image.
In an embodiment of the present invention, the storage module establishes a relational mapping model by using a JSON key-value storage format to form a tree-like data structure that encapsulates and embeds a plurality of the relational mapping models; the data file stored by the storage module is a JSON data file.
In an embodiment of the present invention, the storage module uses the URI to describe a path of the data point set of each tag in the tree data structure.
In an embodiment of the invention, the storage module includes a distributed file storage unit that distributively stores the different types of relational mapping models including the three-dimensional point cloud images and the expansion information thereof into different files and/or distributed database storage units in different database systems.
In an embodiment of the invention, the extension information includes a color, a category label, a category name, a scene category total number, and a scene name.
As described above, the distributed structured three-dimensional point cloud image processing method and system of the present invention have the following beneficial effects:
the method can efficiently and consistently organize and store the three-dimensional point cloud image original data and the structured extension annotation information, and can efficiently perform concurrent access on the structured data model. The present invention is a persistent storage and distribution scheme that supports flexibility and scalability.
Drawings
Fig. 1 is a flowchart illustrating a distributed structured three-dimensional point cloud image processing method according to the present invention.
Fig. 2 is a diagram showing a description of a scene key-value model tree structure in a distributed structured three-dimensional point cloud image processing method according to the present invention.
Fig. 3 is a schematic diagram showing concurrent access in a distributed structured three-dimensional point cloud image processing method according to the present invention.
Fig. 4 is a schematic diagram illustrating a distributed storage in the distributed structured three-dimensional point cloud image processing method according to the present invention.
Fig. 5 is a schematic block diagram of a distributed structured three-dimensional point cloud image processing system according to the present invention.
FIG. 6 is a schematic diagram showing the comparison between the image I/O performance of a distributed structured three-dimensional point cloud image processing method and system according to the present invention and the image I/O performance of the prior art.
Description of the element reference numerals
100 distributed structured three-dimensional point cloud image processing system
101 memory module
102 processing module
S101 to S105
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.
The invention aims to provide a distributed structured three-dimensional point cloud image processing method and a distributed structured three-dimensional point cloud image processing system, which are used for efficiently organizing a structured three-dimensional point cloud image with class label marking data. The principle and the implementation of the distributed structured three-dimensional point cloud image processing method and system of the present invention will be described in detail below, so that those skilled in the art can understand the distributed structured three-dimensional point cloud image processing method and system of the present invention without creative work.
The invention provides a distributed structured three-dimensional point cloud image processing method and a distributed structured three-dimensional point cloud image processing system, in particular to a formatting framework for an input/output task of processing geometric information, color information, coding information and semantic information of a three-dimensional space by adopting a key-value pair data structure, which is used for storing a three-dimensional point cloud image containing semantic annotation or processed by an annotation algorithm.
Referring to fig. 1, a flow chart of a distributed structured three-dimensional point cloud image processing method according to the present invention is shown. As shown in fig. 1, the method of the present invention specifically includes the following steps:
step S101, acquiring a three-dimensional point cloud image and expansion information corresponding to the three-dimensional point cloud image.
In this embodiment, the extended information includes, but is not limited to, a color, a category label, a category name, a total number of scene categories, and a scene name.
Loading the existing three-dimensional point cloud image (including point cloud storage files such as ply and pcd) or the calculated three-dimensional point cloud image into a memory, and in step S101, mainly reading the three-dimensional point cloud image according to the storage mode of a specific point cloud storage file, wherein the three-dimensional point cloud image at least comprises three european space coordinate values, and possibly color information (RGB three channels or RGBA four channels), normal vectors, curvatures and other extension information.
And S102, classifying the acquired three-dimensional point cloud image and the expansion information thereof, establishing a relational mapping model, and forming a tree-shaped data structure for storing and nesting a plurality of relational mapping models.
Classifying all original data points (three-dimensional point cloud images) and expansion information thereof through the relationship among the original data points, establishing different characteristics of point groups, establishing a key-value pair relationship mapping model, and organizing all data into a tree-shaped data structure containing a nested mapping relationship as shown in fig. 2.
In this embodiment, the data points of the three-dimensional point cloud image and the category label object relation map are automatically generated.
The standardized point cloud storage formats currently used in the industry include the + ply and + pcd storage schemes, etc., which all use a single data point as a minimum storage unit (fig. 1), and are arranged in a file by rows, and each data point at least contains three euclidean space coordinate values. For the category label of each data point, numerical value codes are usually used as label codes after manual labeling or automatic labeling. Besides, a single data point may have color information (RGB three-channel or RGBA four-channel), normal vector, curvature, and other extension information, and the dimension of the extension information is expected to be equivalent to the geometric space dimension. The simplest scheme for storing the extended information including the category label is to add the extended information of each data point to the geometric coordinate value and then store and read the extended information (fig. 1), which causes a lot of data redundancy, and the disorder of the original data is not essentially solved.
In the embodiment of the invention, a JSON key-value storage format is adopted to establish a relational mapping model, and a tree-shaped data structure which is embedded and sleeved with a plurality of relational mapping models is formed.
Specifically, the present embodiment adopts the JSON key-value storage format specified in reference to RFC7159 to organize the three-dimensional point cloud image data and the accompanying structured information thereof. For simplifying the description, the expansion information required to be included in the three-dimensional point cloud image data includes: color (RGB), category label, category name, total number of scene categories, scene name. The organization format is specified as follows:
the above organization format establishes a key-value pair relational mapping model by using the relations between all original data points and their extended information, organizes all data into a tree-like data structure containing nested mapping relations, as shown in fig. 2, the model contains both the original data point geometric information and the structured extended information, and removes the redundancy of the extended information. The OBJECT represents an OBJECT type, a plurality of key-value pair relation models are nested inside, STRING is a text STRING type, NUMBER is a numerical value type, ARRAY is a list type, the model can be used for storing three RGB color values, and a nested list can be used for storing all data point sets under a certain class of labels.
Step S103, storing different relational mapping models into corresponding data files through corresponding processing threads for loading and reading. The data file is a JSON data file. And storing different structural models as corresponding JSON files through respective threads for next loading and reading.
Step S104, when a data reading request of the three-dimensional point cloud image is received, establishing a corresponding thread for each stored data file and loading each data file through the corresponding thread to generate corresponding structured data.
That is, when the calculation result is to be used next time, the threads with the same number as the structured data (JSON file) generated in step 103 are first established, and each JSON file is loaded into the memory through different threads.
Step S105, semantic scene category information of a scene corresponding to the three-dimensional point cloud image is obtained according to the structured data, all categories and label codes contained in the scene are analyzed, and the three-dimensional point cloud image is visualized.
In the structured data generated in step S104, scene category information related to semantic meaning of the scene may be acquired from the tag code-tag name corresponding to the codebook key, all categories included in the scene and tag codes thereof are analyzed, and the point cloud is visualized.
Specifically, in the structured data generated in step S104, as shown in fig. 3, scene category information related to semantic meaning of the scene may be obtained from the tag code-tag name corresponding to the codebook key, all categories included in the scene and tag codes thereof are analyzed, and according to the tag codes, all data point sets under the tag category may be queried from the data key. The query operation for the data point set under a single category is independent from the unrelated categories, so that the efficiency of the query operation is ensured, and meanwhile, the data point sets under the labels of a plurality of categories can be accessed simultaneously very easily from the data concurrency perspective. For the concurrent security problem of a single tag, a mutual exclusion variable protection mechanism can be respectively set for the data of each tag to ensure the data security problem under the application of multiple concurrent accesses.
The structured scene key-value model is usually required to be stored persistently, so that an efficient storage scheme with friendly memory and convenient interface is required. One of the simplest solutions is to text the key-value model and store it in a single local file, which has the advantages of simple implementation and convenient operation, and the disadvantages of inconvenient data expansion and insufficient utilization of the structural characteristics of the point data set. Since the data point sets under the data keys in the structured scene key-value model are organized according to tag categories, the data point sets of different categories can be distributively stored to different files, even different physical devices. Because the main data in the key-value model is distributed in the data points of each category under the data key, only the data point set of each category needs to be stored in a distributed manner, and the value corresponding to the label code in the original key-value model is replaced by the unique path identifier of the storage device. One solution is to store data point sets of different labels in multiple files, and the other solution is to distribute data point sets of different labels in a database system, or to store data point sets by using files and databases in a mixed manner, as shown in fig. 4.
In this embodiment, the different types of relational mapping models including the three-dimensional point cloud images and the expansion information thereof are distributively stored in different files and/or different database systems.
In order to establish a Uniform path identifier, the data point set path of each label is described with reference to a URI (Uniform resource identifier) specified in RFC 3986 (character string for identifying and locating any resource). In this embodiment, the path of the data point set of each tag in the tree data structure is described by using a URI.
After the distributed storage operation is performed, the key-value model data only contains non-redundant extension information and path URIs of the data point sets, the file size is usually only a few KB, sharing and distribution are very convenient, and the data point sets are effective for users of the distribution terminals because the URIs are cross-platform.
In addition, as shown in fig. 5, the present embodiment further provides a distributed structured three-dimensional point cloud image processing system 100, where the distributed structured three-dimensional point cloud image processing system 100 includes: a storage module 101 and a processing module 102.
In this embodiment, the storage module 101 is configured to obtain a three-dimensional point cloud image and extension information corresponding to the three-dimensional point cloud image, classify the obtained three-dimensional point cloud image and the extension information thereof, establish a relational mapping model, form a tree-like data structure storing a plurality of nested relational mapping models, and store different relational mapping models as corresponding data files for loading and reading through corresponding processing threads.
In this embodiment, the extension information includes a color, a category label, a category name, a total number of scene categories, and a scene name.
The storage module 101 loads the existing three-dimensional point cloud image (including point cloud storage files such as ply and pcd) or the calculated three-dimensional point cloud image into a memory, and the storage module 101 mainly reads the three-dimensional point cloud image according to the storage mode of a specific point cloud storage file, and at least comprises three Euclidean space coordinate values and possibly expanding information such as color information (RGB three channels or RGBA four channels), normal vectors, curvatures and the like.
The storage module 101 classifies all original data points (three-dimensional point cloud images) and their expansion information according to their relationships, establishes different characteristics to which the point groups belong, establishes a key-value pair relationship mapping model, and organizes all data into a tree-like data structure containing a nested mapping relationship as shown in fig. 2.
In this embodiment, the relational mapping between the data points of the three-dimensional point cloud image and the class label objects is automatically generated, and the storage module 101 establishes a relational mapping model by using a JSON key-value storage format to form a tree-like data structure in which a plurality of relational mapping models are embedded.
Specifically, the present embodiment adopts the JSON key-value storage format specified in reference to RFC7159 to organize the three-dimensional point cloud image data and the accompanying structured information thereof. For simplifying the description, the expansion information required to be included in the three-dimensional point cloud image data includes: color (RGB), category label, category name, total number of scene categories, scene name. The organization format is specified as follows:
the above organization format establishes a key-value pair relational mapping model by using the relations between all original data points and their extended information, organizes all data into a tree-like data structure containing nested mapping relations, as shown in fig. 2, the model contains both the original data point geometric information and the structured extended information, and removes the redundancy of the extended information. The OBJECT represents an OBJECT type, a plurality of key-value pair relation models are nested inside, STRING is a text STRING type, NUMBER is a numerical value type, ARRAY is a list type, the model can be used for storing three RGB color values, and a nested list can be used for storing all data point sets under a certain class of labels.
The data file stored by the storage module 101 is a JSON data file. The storage module 101 stores different structured models as corresponding JSON files for next loading and reading through respective threads.
The structured scene key-value model is usually required to be stored persistently, so that an efficient storage scheme with friendly memory and convenient interface is required. One of the simplest solutions is to text the key-value model and store it in a single local file, which has the advantages of simple implementation and convenient operation, and the disadvantages of inconvenient data expansion and insufficient utilization of the structural characteristics of the point data set. Since the data point sets under the data keys in the structured scene key-value model are organized according to tag categories, the data point sets of different categories can be distributively stored to different files, even different physical devices. Because the main data in the key-value model is distributed in the data points of each category under the data key, only the data point set of each category needs to be stored in a distributed manner, and the value corresponding to the label code in the original key-value model is replaced by the unique path identifier of the storage device. One solution is to store data point sets of different labels in multiple files, and the other solution is to distribute data point sets of different labels in a database system, or to store data point sets by using files and databases in a mixed manner, as shown in fig. 4.
In this embodiment, the storage module 101 includes a distributed file storage unit that distributively stores the different types of relational mapping models including the three-dimensional point cloud images and the expansion information thereof into different files and/or distributed database storage units in different database systems.
In order to establish a Uniform path identifier, the data point set path of each label is described with reference to a URI (Uniform resource identifier) specified in RFC 3986 (character string for identifying and locating any resource). In this embodiment, the storage module 101 uses a URI to describe a path of a data point set of each tag in the tree data structure. After the distributed storage operation is performed, the key-value model data only contains non-redundant extension information and path URIs of the data point sets, the file size is usually only a few KB, sharing and distribution are very convenient, and the data point sets are effective for users of the distribution terminals because the URIs are cross-platform.
In this embodiment, when receiving a data reading request of a three-dimensional point cloud image, the processing module 102 establishes a corresponding thread for each stored data file, loads each data file through the corresponding thread to generate corresponding structured data, obtains semantic scene category information of a scene corresponding to the three-dimensional point cloud image according to the structured data, analyzes all categories and tag codes included in the scene, and visualizes the three-dimensional point cloud image.
Specifically, in the generated structured data, as shown in fig. 3, semantic scene category information about a scene may be obtained from a tag code-tag name corresponding to a codebook key, all categories included in the scene and tag codes thereof may be analyzed, and according to the tag codes, all data point sets under the tag category may be queried from the data key. The query operation for the data point set under a single category is independent from the unrelated categories, so that the efficiency of the query operation is ensured, and meanwhile, the data point sets under the labels of a plurality of categories can be accessed simultaneously very easily from the data concurrency perspective. For the concurrent security problem of a single tag, a mutual exclusion variable protection mechanism can be respectively set for the data of each tag to ensure the data security problem under the application of multiple concurrent accesses.
In order to better illustrate the effectiveness and practical feasibility of the distributed structured three-dimensional point cloud image processing method and system of the present embodiment, the following describes an implementation process of the distributed structured three-dimensional point cloud image processing method and system of the present embodiment with reference to an experimental process.
The experimental process adopts Weinmann M, Jutzi B, Mall let C.Semantic 3D sceneinterrepresentation, a frame combining optical timing neighbor size selection with novel features [ J ]. ISPRS Annals of the Photogrammetry, Remote Sensing and spatial Information Sciences,2014,2(3):181. the Oakland city scan data set used in the experimental process has 5 scene data, the number of data points of each scene is about 100,000, wherein 3 scenes comprise 7 types of point cloud data of different categories, and 2 scenes comprise 5 types of point cloud data of different categories. The Intel Xeon E3-1226 CPU is used for operating a structured point cloud generating program, the performance records are shown in the table 1, the highest performance can reach 10,000 points/second, the requirement of real-time processing can be met, the program is attached to the tail end of a sensor output system, and the three-dimensional point cloud image with the cost meeting the standard of the method provided by the invention can be generated in real time.
TABLE 1 model Generation Performance
#1 | #2 | #3 | #4 | #5 | |
Single file model | 11.4993s | 7.0933s | 10.9682s | 14.0007s | 11.9214s |
Distributed model | 10.60434s | 6.6147s | 10.0461s | 12.7245s | 10.8208s |
And (3) respectively carrying out multithread concurrent loading on the generated three-dimensional point cloud images of the single file model and the distributed model, loading scene extension information and geometric information of the data points into a C + + standard library container, and recording the performance as shown in table 2.
TABLE 2 model analytical Properties
#1 | #2 | #3 | #4 | #5 | |
Single file model | 56.934s | 35.008s | 53.729s | 68.137s | 58.821s |
Distributed model | 38.302s | 20.795s | 29.568s | 47.179s | 37.592s |
As shown in fig. 6, ASCII represents ASCII point cloud loading (dark color) and visualization (light color) time, Binary point cloud loading and visualization time, and Distributed represents point cloud loading and visualization time used in the method of the present invention, it can be seen that the Distributed structured three-dimensional point cloud image processing method and system provided in this embodiment use a concurrency technique, and the I/O performance is equivalent to that of a Binary point cloud.
In summary, the distributed structured three-dimensional point cloud image processing method and system of the present invention achieve the following beneficial effects:
the method can efficiently and consistently organize and store the three-dimensional point cloud image original data and the structured extension annotation information, and can efficiently perform concurrent access on the structured data model. The present invention is a persistent storage and distribution scheme that supports flexibility and scalability. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (8)
1. A distributed structured three-dimensional point cloud image processing method is characterized by comprising the following steps:
acquiring a three-dimensional point cloud image and expansion information corresponding to the three-dimensional point cloud image;
classifying the acquired three-dimensional point cloud image and the expansion information thereof, establishing a relational mapping model, and forming a tree-shaped data structure for storing and nesting a plurality of mapping relational models; establishing a relational mapping model by adopting a JSON key-value storage format to form a tree-shaped data structure which is embedded and sleeved with a plurality of mapping relational models; the data file is a JSON data file;
storing different mapping relation models as corresponding data files for loading and reading through corresponding processing threads;
when a data reading request of a three-dimensional point cloud image is received, establishing a corresponding thread for each stored data file and loading each data file through the corresponding thread to generate corresponding structured data;
and acquiring semantic scene category information of a scene corresponding to the three-dimensional point cloud image according to the structured data, analyzing all categories and label codes contained in the scene, and visualizing the three-dimensional point cloud image.
2. The distributed structured three-dimensional point cloud image processing method according to claim 1, wherein a path of the data point set of each tag in the tree data structure is described by using a URI.
3. The distributed structured three-dimensional point cloud image processing method according to claim 1, wherein different classes of the mapping relationship models containing three-dimensional point cloud images and their extension information are stored in different files and/or different database systems in a distributed manner.
4. The distributed structured three-dimensional point cloud image processing method according to claim 1, wherein the extended information includes a color, a category label, a category name, a scene category total number, and a scene name.
5. A distributed structured three-dimensional point cloud image processing system, comprising:
the storage module is used for acquiring the three-dimensional point cloud image and the expansion information corresponding to the three-dimensional point cloud image, classifying the acquired three-dimensional point cloud image and the expansion information thereof, establishing a relational mapping model, forming a tree-shaped data structure for storing and nesting a plurality of mapping relational models, and storing different mapping relational models as corresponding data files for loading and reading through corresponding processing threads; the storage module establishes a relational mapping model by adopting a JSON key-value storage format to form a tree-shaped data structure which is embedded and sleeved with a plurality of mapping relational models; the data file stored by the storage module is a JSON data file;
the processing module is used for establishing corresponding threads for each stored data file and loading each data file through the corresponding threads to generate corresponding structured data when receiving a data reading request of the three-dimensional point cloud image, acquiring semantic scene category information of a scene corresponding to the three-dimensional point cloud image according to the structured data, analyzing all categories and label codes contained in the scene, and visualizing the three-dimensional point cloud image.
6. The distributed structured three-dimensional point cloud image processing system according to claim 5, wherein the storage module describes a path of the set of data points for each tag in the tree data structure using a URI.
7. The distributed structured three-dimensional point cloud image processing system according to claim 5 or 6, wherein the storage module comprises a distributed file storage unit that distributively stores the different classes of the mapping relation models containing the three-dimensional point cloud images and their extension information into different files and/or a distributed database storage unit in different database systems.
8. The distributed structured three-dimensional point cloud image processing system according to claim 5, wherein the extended information includes a color, a category label, a category name, a scene category total, and a scene name.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710034669.6A CN106844610B (en) | 2017-01-18 | 2017-01-18 | Distributed structured three-dimensional point cloud image processing method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710034669.6A CN106844610B (en) | 2017-01-18 | 2017-01-18 | Distributed structured three-dimensional point cloud image processing method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106844610A CN106844610A (en) | 2017-06-13 |
CN106844610B true CN106844610B (en) | 2020-03-24 |
Family
ID=59124252
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710034669.6A Active CN106844610B (en) | 2017-01-18 | 2017-01-18 | Distributed structured three-dimensional point cloud image processing method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106844610B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109902434B (en) * | 2019-03-18 | 2023-04-07 | 浙江工业大学 | REST architecture style oriented service data visual modeling and matching method in cloud computing environment |
CN110210398A (en) * | 2019-06-03 | 2019-09-06 | 宁波智能装备研究院有限公司 | A kind of three-dimensional point cloud semantic segmentation mask method |
CN111427867B (en) * | 2020-03-30 | 2023-10-20 | 杭州华望系统科技有限公司 | Model persistence method based on hybrid storage |
CN114598692B (en) * | 2020-12-01 | 2023-03-17 | 腾讯科技(深圳)有限公司 | Point cloud file transmission method, application method, device, equipment and storage medium |
CN114661386A (en) * | 2020-12-22 | 2022-06-24 | 腾讯科技(深圳)有限公司 | Point cloud window presenting method and device, computer readable medium and electronic equipment |
CN115080531B (en) * | 2022-05-16 | 2024-07-23 | 重庆紫光华山智安科技有限公司 | Distributed storage image processing method, system, equipment and medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104750854A (en) * | 2015-04-16 | 2015-07-01 | 武汉海达数云技术有限公司 | Mass three-dimensional laser point cloud compression storage and rapid loading and displaying method |
EP2874097A3 (en) * | 2013-11-19 | 2015-07-29 | Nokia Corporation | Automatic scene parsing |
CN105389850A (en) * | 2015-11-03 | 2016-03-09 | 北京大学(天津滨海)新一代信息技术研究院 | Novel visibility generation method for large-scale three-dimensional scene |
CN106331676A (en) * | 2016-08-31 | 2017-01-11 | 贾岳杭 | Three-dimensional data processing and transmission method based on virtual reality environment |
-
2017
- 2017-01-18 CN CN201710034669.6A patent/CN106844610B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2874097A3 (en) * | 2013-11-19 | 2015-07-29 | Nokia Corporation | Automatic scene parsing |
CN104750854A (en) * | 2015-04-16 | 2015-07-01 | 武汉海达数云技术有限公司 | Mass three-dimensional laser point cloud compression storage and rapid loading and displaying method |
CN105389850A (en) * | 2015-11-03 | 2016-03-09 | 北京大学(天津滨海)新一代信息技术研究院 | Novel visibility generation method for large-scale three-dimensional scene |
CN106331676A (en) * | 2016-08-31 | 2017-01-11 | 贾岳杭 | Three-dimensional data processing and transmission method based on virtual reality environment |
Non-Patent Citations (1)
Title |
---|
LiDAR点云数据的分布式组织及其并行获取方法;王宗跃等;《武汉大学学报 信息科学版》;20090831;第34卷(第8期);第936-939页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106844610A (en) | 2017-06-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106844610B (en) | Distributed structured three-dimensional point cloud image processing method and system | |
CN102682103B (en) | Three-dimensional space index method aiming at massive laser radar point cloud models | |
US9396512B2 (en) | Fully parallel construction of k-d trees, octrees, and quadtrees in a graphics processing unit | |
Wang et al. | Trinary-projection trees for approximate nearest neighbor search | |
Najdataei et al. | Continuous and parallel lidar point-cloud clustering | |
CN101673410A (en) | Vector building drawing based method for reconstructing three-dimensional model | |
Schön et al. | Octree-based indexing for 3D pointclouds within an Oracle Spatial DBMS | |
CN113535788B (en) | Ocean environment data-oriented retrieval method, system, equipment and medium | |
Puri et al. | MapReduce algorithms for GIS polygonal overlay processing | |
CN114385620A (en) | Data processing method, device, equipment and readable storage medium | |
Krahe et al. | AI-based Computer Aided Engineering for automated product design-A first approach with a Multi-View based classification | |
CN111611448A (en) | Knowledge-driven joint big data query and analysis platform | |
Singh et al. | Spatial data analysis with ArcGIS and MapReduce | |
Nguyen et al. | Spatio-semantic comparison of large 3D city models in CityGML using a graph database | |
CN115905630A (en) | Graph database query method, device, equipment and storage medium | |
Kim et al. | Utilizing extended geocodes for handling massive three-dimensional point cloud data | |
US20110191359A1 (en) | Process for transforming and consulting directed and attributed multigraphs based on the use of maps and bitmaps | |
Jouili et al. | imGraph: A distributed in-memory graph database | |
Ojal et al. | A universal method to compare parts from STEP files | |
Uher et al. | Searching of circular neighborhoods in the square and hexagonal regular grids | |
CN116226686B (en) | Table similarity analysis method, apparatus, device and storage medium | |
CN117540056B (en) | Method, device, computer equipment and storage medium for data query | |
KR102596257B1 (en) | Method for recognizing machining feature and computing device for executing the method | |
CN116842740B (en) | Intelligent identification method for processing characteristics based on primitive model | |
Fontolan | Modularity based community detection on the GPU |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |