CN112052884A - Point cloud classification method and system based on local edge feature enhancement - Google Patents
Point cloud classification method and system based on local edge feature enhancement Download PDFInfo
- Publication number
- CN112052884A CN112052884A CN202010827732.3A CN202010827732A CN112052884A CN 112052884 A CN112052884 A CN 112052884A CN 202010827732 A CN202010827732 A CN 202010827732A CN 112052884 A CN112052884 A CN 112052884A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- data
- module
- feature
- classification
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to the technical field of point cloud classification, and discloses a point cloud classification method and system based on local edge feature enhancement, wherein the point cloud classification system based on local edge feature enhancement comprises: the system comprises a point cloud data acquisition module, a data preprocessing module, a main control module, an edge feature extraction module, an edge feature enhancement module, a feature filling module, a point cloud classification model construction module, a point cloud classification module, a cloud storage module and a display module. According to the invention, point cloud voxelization data, edge features of point clouds corresponding to the point cloud voxelization data in a preset neighborhood and voxel positions corresponding to the point clouds are obtained, a point cloud classification model is constructed based on a graph convolution network structure and a channel attention mechanism to classify the point clouds after feature filling, and a point cloud classification result is output, so that the interdependence relationship among feature channels is increased, the global feature representation of the point clouds is enhanced, and the efficiency and the prediction accuracy of point cloud classification are improved.
Description
Technical Field
The invention belongs to the technical field of point cloud classification, and particularly relates to a point cloud classification method and system based on local edge feature enhancement.
Background
At present, since the development of deep learning technology, researchers begin to explore how to utilize a deep neural network to perform point cloud feature learning, and three-dimensional point cloud classification and segmentation based on deep learning also become more and more important directions for researchers.
The traditional point cloud classification method mainly utilizes the domain knowledge of people to manually construct features, and uses algorithms such as a Support Vector Machine (SVM), a Random Forest (RF) and the like to classify the manually constructed features extracted from a sample. The method needs to design manual characteristics aiming at different tasks and different data and relying on rich practical experience, thereby not only consuming manpower, but also having limitation on the expression capability of the characteristics. In addition, the traditional point cloud classification method has low point cloud classification accuracy, cannot enhance the edge characteristics and cannot meet the existing requirements. Therefore, a new point cloud classification method based on local edge feature enhancement is needed.
Through the above analysis, the problems and defects of the prior art are as follows:
(1) the traditional point cloud classification method needs to design manual features aiming at different tasks, different data and rich practical experience, not only consumes manpower, but also has limitation on the expression capability of the features.
(2) The traditional point cloud classification method has low point cloud classification accuracy, cannot enhance edge characteristics and cannot meet the existing requirements.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a point cloud classification method and system based on local edge feature enhancement.
The invention is realized in such a way that a point cloud classification method based on local edge feature enhancement comprises the following steps:
the method comprises the steps that firstly, a point cloud data acquisition module receives a point cloud classification request through data acquisition equipment, and point cloud data in the point cloud classification request are acquired;
performing point cloud voxelization operation on the point cloud corresponding to the point cloud data by using a data preprocessing program through a data preprocessing module to obtain point cloud voxelization data;
the method for performing point cloud voxelization operation on point clouds corresponding to the point cloud data by using a data preprocessing program through a data preprocessing module comprises the following steps of:
(2.1) converting the acquired point cloud data into a TXT text document format of a Cartesian coordinate system, and representing three-dimensional point cloud data in the form of X, Y, Z Cartesian coordinates;
(2.2) X, Y, Z Cartesian coordinate system based on three-dimensional point cloud data and (X)min，Ymin，Zmin) Determining the voxel coordinate of the point cloud data in the space according to an rounding algorithm, and converting the single point cloud data into a three-dimensional voxel value;
controlling the normal operation of each module of the point cloud classification system based on local edge feature enhancement by using a main control module and a main controller;
acquiring edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood by using a feature extraction program through an edge feature extraction module;
the method for acquiring the edge features of the point cloud corresponding to the point cloud voxelization data in the preset neighborhood comprises the following steps:
(4.1) acquiring point cloud voxelization data, and processing the acquired point cloud voxelization data to obtain point cloud distribution data;
(4.2) processing the obtained point cloud distribution data to obtain associated characteristic distance data and point cloud vector data of the corresponding point cloud in a preset neighborhood;
(4.3) acquiring point cloud similar data according to the obtained associated feature distance data;
(4.4) classifying the point cloud vector data according to the obtained point cloud similar data to obtain the edge characteristics of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood;
fifthly, utilizing a feature enhancement program to enhance the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through an edge feature enhancement module;
step six, acquiring a voxel position corresponding to the point cloud voxel data by a feature filling module through a feature filling program, and filling the edge feature after the enhancement processing in the voxel position corresponding to the point cloud; constructing a point cloud classification model by using a model construction program through a point cloud classification model construction module;
classifying the point cloud filled with the edge features by using a point cloud classification program through a point cloud classification module and a constructed cloud classification model, and outputting a point cloud classification result corresponding to the point cloud data;
step eight, storing the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the point cloud classification result by using a cloud database server through a cloud storage module;
and step nine, displaying the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the real-time data of the point cloud classification result by using an LED high-definition display through a display module.
Further, in the step (2.2), the step length is 2 cm.
Further, in the step (2.2), the converting the single point cloud data into three-dimensional voxel values includes:
wherein i, j, k are corresponding voxel coordinates, and L is a step size, i.e., a voxel value size.
Further, in the seventh step, the method for classifying the point cloud after the edge feature filling by using the point cloud classification module and the established cloud classification program comprises:
(1) inputting point cloud data into the constructed cloud classification model; wherein the point cloud classification model comprises: at least one KNN map convolution element and a channel attention element;
(2) sequentially extracting feature aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through each KNN image convolution unit;
(3) sequentially subjecting the feature aggregation data to multilayer perceptron and pooling to obtain a first global point cloud feature; -
(4) Inputting the first global point cloud feature into the channel attention unit, and obtaining an adjusted second global point cloud feature through the channel attention unit;
(5) and obtaining a point cloud classification result according to the adjusted second global point cloud characteristic.
Further, the KNN graph convolution unit comprises an edge feature extraction subunit and a feature aggregation subunit.
Further, the method for extracting feature aggregation data of all edges of each vertex in the point cloud data included in the point cloud data sequentially through the KNN map convolution modules comprises the following steps:
1) inputting the central node features, neighborhood node features and feature difference values among the nodes of each point data in the point cloud data into the feature extraction subunit, and obtaining the edge features connected among the point clouds in the point cloud data through the feature extraction subunit;
2) and inputting the edge features into a feature aggregation subunit, and obtaining feature aggregation data of all edges of each vertex in the point cloud data through the feature aggregation subunit.
Another object of the present invention is to provide a local edge feature enhancement-based point cloud classification system implementing the local edge feature enhancement-based point cloud classification method, the local edge feature enhancement-based point cloud classification system including:
the system comprises a point cloud data acquisition module, a data preprocessing module, a main control module, an edge feature extraction module, an edge feature enhancement module, a feature filling module, a point cloud classification model construction module, a point cloud classification module, a cloud storage module and a display module.
The point cloud data acquisition module is connected with the main control module and used for receiving a point cloud classification request through data acquisition equipment and acquiring point cloud data in the point cloud classification request;
the data preprocessing module is connected with the main control module and used for carrying out point cloud voxelization operation on the point cloud corresponding to the point cloud data through a data preprocessing program to obtain point cloud voxelization data;
the main control module is connected with the point cloud data acquisition module, the data preprocessing module, the edge feature extraction module, the edge feature enhancement module, the feature filling module, the point cloud classification model construction module, the point cloud classification module, the cloud storage module and the display module and is used for controlling the normal operation of each module of the point cloud classification system based on local edge feature enhancement through the main controller;
the edge feature extraction module is connected with the main control module and used for acquiring the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through a feature extraction program;
the edge feature enhancement module is connected with the main control module and used for enhancing the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through a feature enhancement program;
the characteristic filling module is connected with the main control module and used for acquiring a voxel position corresponding to the point cloud voxelization data through a characteristic filling program and filling the edge characteristic after the enhancement processing in the voxel position corresponding to the point cloud;
the point cloud classification model building module is connected with the main control module and used for building a point cloud classification model through a model building program;
the point cloud classification module is connected with the main control module and used for classifying the point cloud filled with the edge features by utilizing the constructed cloud classification model through a point cloud classification program and outputting a point cloud classification result corresponding to the point cloud data;
the cloud storage module is connected with the main control module and used for storing the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the point cloud classification result through a cloud database server;
and the display module is connected with the main control module and used for displaying the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the real-time data of the point cloud classification result through an LED high-definition display.
Further, the point cloud classification module includes:
the data input unit is used for inputting point cloud data into the constructed cloud classification model; the point cloud classification model comprises: at least one KNN map convolution element and a channel attention element;
the characteristic aggregation data extraction unit is used for extracting characteristic aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through the KNN graph convolution units;
the first global point cloud feature acquisition unit is used for sequentially carrying out multilayer perceptron and pooling on the feature aggregation data to obtain a first global point cloud feature;
the second global point cloud feature acquisition unit is used for inputting the first global point cloud feature into the channel attention unit and obtaining the adjusted second global point cloud feature through the channel attention unit;
and the classification result output unit is used for obtaining a point cloud classification result according to the adjusted second global point cloud characteristic and outputting and displaying the point cloud classification result.
It is another object of the present invention to provide a computer program product stored on a computer readable medium, comprising a computer readable program for providing a user input interface to implement the method for point cloud classification based on local edge feature enhancement when executed on an electronic device.
Another object of the present invention is to provide a computer-readable storage medium storing instructions which, when executed on a computer, cause the computer to execute the method for point cloud classification based on local edge feature enhancement.
By combining all the technical schemes, the invention has the advantages and positive effects that: the point cloud classification method based on local edge feature enhancement provided by the invention comprises the steps of receiving a point cloud classification request, obtaining point cloud data in the point cloud classification request, carrying out point cloud voxelization operation on point clouds corresponding to the point cloud data to obtain point cloud voxelization data, obtaining edge features of the point clouds corresponding to the point cloud voxelization data in a preset neighborhood, obtaining voxel positions corresponding to the point clouds corresponding to the point cloud voxelization data, filling the obtained edge features in the voxel positions corresponding to the point clouds, constructing a point cloud classification model based on a graph volume network structure and a channel attention mechanism to classify the point clouds after feature filling, outputting a point cloud classification result corresponding to the point cloud data, increasing the interdependency relation among feature channels, strengthening the global feature representation of the point clouds, and improving the efficiency and the prediction accuracy of point cloud classification.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained from the drawings without creative efforts.
FIG. 1 is a block diagram of a point cloud classification system based on local edge feature enhancement according to an embodiment of the present invention;
in the figure: 1. a point cloud data acquisition module; 2. a data preprocessing module; 3. a main control module; 4. an edge feature extraction module; 5. an edge feature enhancement module; 6. a feature filling module; 7. a point cloud classification model building module; 8. a point cloud classification module; 9. a cloud storage module; 10. and a display module.
Fig. 2 is a flowchart of a point cloud classification method based on local edge feature enhancement according to an embodiment of the present invention.
Fig. 3 is a flowchart of a method for performing point cloud voxelization operation on a point cloud corresponding to the point cloud data by using a data preprocessing program through a data preprocessing module according to an embodiment of the present invention.
Fig. 4 is a flowchart of a method for obtaining edge features of a point cloud in a preset neighborhood corresponding to the point cloud voxelization data according to the embodiment of the present invention.
Fig. 5 is a flowchart of a method for classifying the point cloud after the edge feature is filled by using the point cloud classification module and the established cloud classification model through the point cloud classification program according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides a point cloud classification method and system based on local edge feature enhancement, and the invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the point cloud classification method based on local edge feature enhancement provided by the embodiment of the present invention includes the following steps:
s101, receiving a point cloud classification request by a point cloud data acquisition module through data acquisition equipment, and acquiring point cloud data in the point cloud classification request.
And S102, performing point cloud voxelization operation on the point cloud corresponding to the point cloud data by using a data preprocessing program through a data preprocessing module to obtain point cloud voxelization data.
S103, controlling the normal operation of each module of the point cloud classification system based on local edge feature enhancement by using a main control module and a main controller.
And S104, acquiring edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood by using a feature extraction program through an edge feature extraction module.
And S105, utilizing a feature enhancement program to enhance the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through an edge feature enhancement module.
S106, acquiring a voxel position corresponding to the point cloud voxelization data by a feature filling module through a feature filling program, and filling the edge feature after the enhancement processing in the voxel position corresponding to the point cloud; and constructing a point cloud classification model by using a model construction program through a point cloud classification model construction module.
And S107, classifying the point cloud filled with the edge features by using a point cloud classification program through a point cloud classification module, and outputting a point cloud classification result corresponding to the point cloud data.
And S108, storing the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the point cloud classification result by using a cloud database server through a cloud storage module.
And S109, displaying the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the real-time data of the point cloud classification result by using an LED high-definition display through a display module.
As shown in fig. 2, the point cloud classification system based on local edge feature enhancement provided by the embodiment of the present invention includes: the system comprises a point cloud data acquisition module 1, a data preprocessing module 2, a main control module 3, an edge feature extraction module 4, an edge feature enhancement module 5, a feature filling module 6, a point cloud classification model building module 7, a point cloud classification module 8, a cloud storage module 9 and a display module 10.
The point cloud data acquisition module 1 is connected with the main control module 3 and used for receiving a point cloud classification request through data acquisition equipment and acquiring point cloud data in the point cloud classification request;
the data preprocessing module 2 is connected with the main control module 3 and is used for carrying out point cloud voxelization operation on the point cloud corresponding to the point cloud data through a data preprocessing program to obtain point cloud voxelization data;
the main control module 3 is connected with the point cloud data acquisition module 1, the data preprocessing module 2, the edge feature extraction module 4, the edge feature enhancement module 5, the feature filling module 6, the point cloud classification model construction module 7, the point cloud classification module 8, the cloud storage module 9 and the display module 10, and is used for controlling the normal operation of each module of the point cloud classification system based on local edge feature enhancement through the main controller;
the edge feature extraction module 4 is connected with the main control module 3 and is used for acquiring the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through a feature extraction program;
the edge feature enhancement module 5 is connected with the main control module 3 and is used for enhancing the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through a feature enhancement program;
the characteristic filling module 6 is connected with the main control module 3 and used for acquiring a voxel position corresponding to the point cloud voxelization data through a characteristic filling program and filling the edge characteristic after the enhancement processing in the voxel position corresponding to the point cloud;
the point cloud classification model building module 7 is connected with the main control module 3 and used for building a point cloud classification model through a model building program;
the point cloud classification module 8 is connected with the main control module 3 and used for classifying the point cloud filled with the edge features by utilizing the constructed cloud classification model through a point cloud classification program and outputting a point cloud classification result corresponding to the point cloud data;
the cloud storage module 9 is connected with the main control module 3 and used for storing the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the point cloud classification result through a cloud database server;
and the display module 10 is connected with the main control module 3 and is used for displaying the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the real-time data of the point cloud classification result through an LED high-definition display.
The point cloud classification module 8 provided by the embodiment of the invention comprises:
the data input module 8-1 is used for inputting point cloud data into the constructed cloud classification model; the point cloud classification model comprises: at least one KNN map convolution element and a channel attention element;
the feature aggregation data extraction module 8-2 is used for extracting feature aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through each KNN graph convolution unit;
the first global point cloud feature acquisition module 8-3 is used for sequentially subjecting the feature aggregation data to multilayer perceptron and pooling to obtain a first global point cloud feature;
a second global point cloud feature obtaining module 8-4, configured to input the first global point cloud feature into the channel attention unit, and obtain an adjusted second global point cloud feature through the channel attention unit;
and the classification result output module 8-5 is used for obtaining a point cloud classification result according to the adjusted second global point cloud characteristics and outputting and displaying the point cloud classification result.
The technical solution of the present invention is further illustrated by the following specific examples.
Example 1:
the point cloud classification method based on local edge feature enhancement provided by the embodiment of the invention is shown in fig. 1, and as a preferred embodiment, as shown in fig. 3, the method for performing point cloud voxelization operation on a point cloud corresponding to point cloud data by using a data preprocessing program through a data preprocessing module provided by the embodiment of the invention comprises the following steps:
s201, converting the acquired point cloud data into a TXT text document format of a Cartesian coordinate system, and representing three-dimensional point cloud data by the Cartesian coordinates in the form of X, Y, Z.
S202, X, Y, Z Cartesian coordinate system based on three-dimensional point cloud data is represented by (X)min，Ymin，Zmin) And determining the voxel coordinate of the point cloud data in the space according to an rounding algorithm, and converting the single point cloud data into a three-dimensional voxel value.
In step S202, the step length provided by the embodiment of the present invention is 2 cm.
In step S202, the converting of the single point cloud data into the three-dimensional voxel value according to the embodiment of the present invention includes:
wherein i, j, k are corresponding voxel coordinates, and L is a step size, i.e., a voxel value size.
Example 2:
the point cloud classification method based on local edge feature enhancement provided by the embodiment of the invention is shown in fig. 1, and as a preferred embodiment, as shown in fig. 4, the method for obtaining the edge feature of the point cloud corresponding to the point cloud voxelization data in the preset neighborhood provided by the embodiment of the invention comprises the following steps:
s301, acquiring point cloud voxelization data, and processing the acquired point cloud voxelization data to obtain point cloud distribution data;
s302, processing the obtained point cloud distribution data to obtain associated characteristic distance data and point cloud vector data of the corresponding point cloud in a preset neighborhood;
s303, acquiring point cloud similar data according to the obtained associated feature distance data;
s304, classifying the point cloud vector data according to the obtained point cloud similar data to obtain the edge features of the point cloud corresponding to the point cloud voxelization data in the preset neighborhood.
Example 3:
as shown in fig. 1, and as shown in fig. 5, the method for classifying a point cloud filled with edge features by using a built cloud classification model through a point cloud classification module and using a point cloud classification program according to an embodiment of the present invention includes:
s401, inputting point cloud data into the constructed cloud classification model; wherein the point cloud classification model comprises: at least one KNN map convolution element and a channel attention element.
S402, sequentially extracting feature aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through the KNN image convolution units.
And S403, sequentially subjecting the feature aggregation data to multilayer perceptron and pooling to obtain a first global point cloud feature.
S404, inputting the first global point cloud feature into the channel attention unit, and obtaining an adjusted second global point cloud feature through the channel attention unit.
And S405, obtaining a point cloud classification result according to the adjusted second global point cloud feature.
The KNN graph convolution unit provided by the embodiment of the invention comprises an edge feature extraction subunit and a feature aggregation subunit; the method for extracting feature aggregation data of all edges of each vertex in point cloud data contained in the point cloud data sequentially through the KNN image convolution modules comprises the following steps:
1) inputting the central node features, neighborhood node features and feature difference values among the nodes of each point data in the point cloud data into the feature extraction subunit, and obtaining the edge features connected among the point clouds in the point cloud data through the feature extraction subunit;
2) and inputting the edge features into a feature aggregation subunit, and obtaining feature aggregation data of all edges of each vertex in the point cloud data through the feature aggregation subunit.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When used in whole or in part, can be implemented in a computer program product that includes one or more computer instructions. When loaded or executed on a computer, cause the flow or functions according to embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center via wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL), or wireless (e.g., infrared, wireless, microwave, etc.)). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that includes one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
The above description is only for the purpose of illustrating the present invention and the appended claims are not to be construed as limiting the scope of the invention, which is intended to cover all modifications, equivalents and improvements that are within the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A point cloud classification method based on local edge feature enhancement is characterized by comprising the following steps:
the method comprises the steps that firstly, a point cloud data acquisition module receives a point cloud classification request through data acquisition equipment, and point cloud data in the point cloud classification request are acquired;
performing point cloud voxelization operation on the point cloud corresponding to the point cloud data by using a data preprocessing program through a data preprocessing module to obtain point cloud voxelization data;
the method for performing point cloud voxelization operation on point clouds corresponding to the point cloud data by using a data preprocessing program through a data preprocessing module comprises the following steps of:
(2.1) converting the acquired point cloud data into a TXT text document format of a Cartesian coordinate system, and representing three-dimensional point cloud data in the form of X, Y, Z Cartesian coordinates;
(2.2) X, Y, Z Cartesian coordinate system based on three-dimensional point cloud data and (X)min，Ymin，Zmin) Determining the voxel coordinate of the point cloud data in the space according to an rounding algorithm, and converting the single point cloud data into a three-dimensional voxel value;
controlling the normal operation of each module of the point cloud classification system based on local edge feature enhancement by using a main control module and a main controller;
acquiring edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood by using a feature extraction program through an edge feature extraction module;
the method for acquiring the edge features of the point cloud corresponding to the point cloud voxelization data in the preset neighborhood comprises the following steps:
(4.1) acquiring point cloud voxelization data, and processing the acquired point cloud voxelization data to obtain point cloud distribution data;
(4.2) processing the obtained point cloud distribution data to obtain associated characteristic distance data and point cloud vector data of the corresponding point cloud in a preset neighborhood;
(4.3) acquiring point cloud similar data according to the obtained associated feature distance data;
(4.4) classifying the point cloud vector data according to the obtained point cloud similar data to obtain the edge characteristics of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood;
fifthly, utilizing a feature enhancement program to enhance the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through an edge feature enhancement module;
step six, acquiring a voxel position corresponding to the point cloud voxel data by a feature filling module through a feature filling program, and filling the edge feature after the enhancement processing in the voxel position corresponding to the point cloud; constructing a point cloud classification model by using a model construction program through a point cloud classification model construction module;
classifying the point cloud filled with the edge features by using a point cloud classification program through a point cloud classification module and a constructed cloud classification model, and outputting a point cloud classification result corresponding to the point cloud data;
step eight, storing the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the point cloud classification result by using a cloud database server through a cloud storage module;
and step nine, displaying the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the real-time data of the point cloud classification result by using an LED high-definition display through a display module.
2. The point cloud classification method based on local edge feature enhancement as claimed in claim 1, wherein in step (2.2), the step size is 2 cm.
3. The local edge feature enhancement-based point cloud classification method according to claim 1, wherein in the step (2.2), the converting the single point cloud data into three-dimensional voxel values comprises:
wherein i, j, k are corresponding voxel coordinates, and L is a step size, i.e., a voxel value size.
4. The method for classifying point cloud based on local edge feature enhancement according to claim 1, wherein in step seven, the method for classifying the point cloud after the edge feature filling by using the point cloud classification module and the point cloud classification program and the constructed cloud classification model comprises:
(1) inputting point cloud data into the constructed cloud classification model; wherein the point cloud classification model comprises: at least one KNN map convolution element and a channel attention element;
(2) sequentially extracting feature aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through each KNN image convolution unit;
(3) sequentially subjecting the feature aggregation data to multilayer perceptron and pooling to obtain a first global point cloud feature;
(4) inputting the first global point cloud feature into the channel attention unit, and obtaining an adjusted second global point cloud feature through the channel attention unit;
(5) and obtaining a point cloud classification result according to the adjusted second global point cloud characteristic.
5. The point cloud classification method based on local edge feature enhancement as claimed in claim 4, wherein the KNN map convolution unit comprises an edge feature extraction subunit and a feature aggregation subunit.
6. The point cloud classification method based on local edge feature enhancement according to claim 4, wherein the method for sequentially extracting feature aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through the KNN image volume modules comprises the following steps:
1) inputting the central node features, neighborhood node features and feature difference values among the nodes of each point data in the point cloud data into the feature extraction subunit, and obtaining the edge features connected among the point clouds in the point cloud data through the feature extraction subunit;
2) and inputting the edge features into a feature aggregation subunit, and obtaining feature aggregation data of all edges of each vertex in the point cloud data through the feature aggregation subunit.
7. A local edge feature enhancement based point cloud classification system implementing the local edge feature enhancement based point cloud classification method according to any one of claims 1 to 6, wherein the local edge feature enhancement based point cloud classification system comprises:
the system comprises a point cloud data acquisition module, a data preprocessing module, a main control module, an edge feature extraction module, an edge feature enhancement module, a feature filling module, a point cloud classification model construction module, a point cloud classification module, a cloud storage module and a display module;
the point cloud data acquisition module is connected with the main control module and used for receiving a point cloud classification request through data acquisition equipment and acquiring point cloud data in the point cloud classification request;
the data preprocessing module is connected with the main control module and used for carrying out point cloud voxelization operation on the point cloud corresponding to the point cloud data through a data preprocessing program to obtain point cloud voxelization data;
the main control module is connected with the point cloud data acquisition module, the data preprocessing module, the edge feature extraction module, the edge feature enhancement module, the feature filling module, the point cloud classification model construction module, the point cloud classification module, the cloud storage module and the display module and is used for controlling the normal operation of each module of the point cloud classification system based on local edge feature enhancement through the main controller;
the edge feature extraction module is connected with the main control module and used for acquiring the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through a feature extraction program;
the edge feature enhancement module is connected with the main control module and used for enhancing the edge features of the point cloud corresponding to the point cloud voxelization data in a preset neighborhood through a feature enhancement program;
the characteristic filling module is connected with the main control module and used for acquiring a voxel position corresponding to the point cloud voxelization data through a characteristic filling program and filling the edge characteristic after the enhancement processing in the voxel position corresponding to the point cloud;
the point cloud classification model building module is connected with the main control module and used for building a point cloud classification model through a model building program;
the point cloud classification module is connected with the main control module and used for classifying the point cloud filled with the edge features by utilizing the constructed cloud classification model through a point cloud classification program and outputting a point cloud classification result corresponding to the point cloud data;
the cloud storage module is connected with the main control module and used for storing the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the point cloud classification result through a cloud database server;
and the display module is connected with the main control module and used for displaying the acquired point cloud data, the processed point cloud voxelization data, the edge characteristics, the point cloud classification model and the real-time data of the point cloud classification result through an LED high-definition display.
8. The point cloud classification system based on local edge feature enhancement as claimed in claim 7, wherein the point cloud classification module comprises:
the data input unit is used for inputting point cloud data into the constructed cloud classification model; the point cloud classification model comprises: at least one KNN map convolution element and a channel attention element;
the characteristic aggregation data extraction unit is used for extracting characteristic aggregation data of all edges of each vertex in the point cloud data contained in the point cloud data through the KNN graph convolution units;
the first global point cloud feature acquisition unit is used for sequentially carrying out multilayer perceptron and pooling on the feature aggregation data to obtain a first global point cloud feature;
the second global point cloud feature acquisition unit is used for inputting the first global point cloud feature into the channel attention unit and obtaining the adjusted second global point cloud feature through the channel attention unit;
and the classification result output unit is used for obtaining a point cloud classification result according to the adjusted second global point cloud characteristic and outputting and displaying the point cloud classification result.
9. A computer program product stored on a computer readable medium, comprising a computer readable program for providing a user input interface to implement the method of point cloud classification based on local edge feature enhancement as claimed in any one of claims 3 to 8 when executed on an electronic device.
10. A computer-readable storage medium storing instructions which, when executed on a computer, cause the computer to perform the point cloud classification method based on local edge feature enhancement according to any one of claims 3 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010827732.3A CN112052884A (en) | 2020-08-17 | 2020-08-17 | Point cloud classification method and system based on local edge feature enhancement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010827732.3A CN112052884A (en) | 2020-08-17 | 2020-08-17 | Point cloud classification method and system based on local edge feature enhancement |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112052884A true CN112052884A (en) | 2020-12-08 |
Family
ID=73600059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010827732.3A Pending CN112052884A (en) | 2020-08-17 | 2020-08-17 | Point cloud classification method and system based on local edge feature enhancement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112052884A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112633376A (en) * | 2020-12-24 | 2021-04-09 | 南京信息工程大学 | Point cloud data ground feature classification method and system based on deep learning and storage medium |
| CN113642712A (en) * | 2021-08-17 | 2021-11-12 | 成都视海芯图微电子有限公司 | Point cloud data processor and method based on deep learning |
| CN116128835A (en) * | 2023-02-02 | 2023-05-16 | 南方电网数字电网研究院有限公司 | Point cloud analysis-based power transmission wire sag measurement method, device and equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106933786A (en) * | 2017-03-07 | 2017-07-07 | 福建农林大学 | A kind of three-dimensional laser point cloud data rapid voxel processing method |
| CN109426828A (en) * | 2017-08-28 | 2019-03-05 | 中国科学院深圳先进技术研究院 | Point cloud classifications method, apparatus, equipment and storage medium |
| US20190179024A1 (en) * | 2017-12-13 | 2019-06-13 | Luminar Technologies, Inc. | Processing point clouds of vehicle sensors having variable scan line distributions using voxel grids |
| CN110516751A (en) * | 2019-08-29 | 2019-11-29 | 上海交通大学 | Processing method, system and the equipment of three-dimensional data |
| CN110569926A (en) * | 2019-09-19 | 2019-12-13 | 北京工商大学 | point cloud classification method based on local edge feature enhancement |
| CN111242208A (en) * | 2020-01-08 | 2020-06-05 | 深圳大学 | Point cloud classification method, point cloud segmentation method and related equipment |
-
2020
- 2020-08-17 CN CN202010827732.3A patent/CN112052884A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106933786A (en) * | 2017-03-07 | 2017-07-07 | 福建农林大学 | A kind of three-dimensional laser point cloud data rapid voxel processing method |
| CN109426828A (en) * | 2017-08-28 | 2019-03-05 | 中国科学院深圳先进技术研究院 | Point cloud classifications method, apparatus, equipment and storage medium |
| US20190179024A1 (en) * | 2017-12-13 | 2019-06-13 | Luminar Technologies, Inc. | Processing point clouds of vehicle sensors having variable scan line distributions using voxel grids |
| CN110516751A (en) * | 2019-08-29 | 2019-11-29 | 上海交通大学 | Processing method, system and the equipment of three-dimensional data |
| CN110569926A (en) * | 2019-09-19 | 2019-12-13 | 北京工商大学 | point cloud classification method based on local edge feature enhancement |
| CN111242208A (en) * | 2020-01-08 | 2020-06-05 | 深圳大学 | Point cloud classification method, point cloud segmentation method and related equipment |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112633376A (en) * | 2020-12-24 | 2021-04-09 | 南京信息工程大学 | Point cloud data ground feature classification method and system based on deep learning and storage medium |
| CN113642712A (en) * | 2021-08-17 | 2021-11-12 | 成都视海芯图微电子有限公司 | Point cloud data processor and method based on deep learning |
| CN113642712B (en) * | 2021-08-17 | 2023-08-08 | 成都视海芯图微电子有限公司 | Point cloud data processor and method based on deep learning |
| CN116128835A (en) * | 2023-02-02 | 2023-05-16 | 南方电网数字电网研究院有限公司 | Point cloud analysis-based power transmission wire sag measurement method, device and equipment |
| CN116128835B (en) * | 2023-02-02 | 2023-09-15 | 南方电网数字电网研究院有限公司 | Point cloud analysis-based power transmission wire sag measurement method, device and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112528977B (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN112052884A (en) | Point cloud classification method and system based on local edge feature enhancement | |
| CN111488985B (en) | Deep neural network model compression training method, device, equipment and medium | |
| CN113221743B (en) | Table analysis method, apparatus, electronic device and storage medium | |
| CN114648676B (en) | Training method of point cloud processing model and point cloud instance segmentation method and device | |
| JP7242994B2 (en) | Video event identification method, apparatus, electronic device and storage medium | |
| US20210166058A1 (en) | Image generation method and computing device | |
| CN114792355B (en) | Virtual image generation method and device, electronic equipment and storage medium | |
| CN113902010A (en) | Training method of classification model, image classification method, device, equipment and medium | |
| CN114566233A (en) | Method, device, electronic device and storage medium for molecular screening | |
| CN113162787B (en) | Method for fault location in a telecommunication network, node classification method and related devices | |
| CN112241789A (en) | Structured pruning method, device, medium and equipment for lightweight neural network | |
| WO2023035535A1 (en) | Training method and apparatus for semantic segmentation network, device, and storage medium | |
| CN113762109B (en) | Training method of character positioning model and character positioning method | |
| WO2021000411A1 (en) | Neural network-based document classification method and apparatus, and device and storage medium | |
| CN114861059A (en) | Resource recommendation method and device, electronic equipment and storage medium | |
| US11881044B2 (en) | Method and apparatus for processing image, device and storage medium | |
| CN112784102B (en) | Video retrieval method and device and electronic equipment | |
| CN111709473B (en) | Clustering method and device for object features | |
| CN113190551A (en) | Feature retrieval system construction method, feature retrieval method, device and equipment | |
| CN114037772A (en) | Training method of image generator, image generation method and device | |
| CN117971487A (en) | High-performance operator generation method, device, equipment and storage medium | |
| CN113344213A (en) | Knowledge distillation method, knowledge distillation device, electronic equipment and computer readable storage medium | |
| CN113139110A (en) | Regional feature processing method, device, equipment, storage medium and program product | |
| CN115495248B (en) | Memory allocation method and device of reasoning card, electronic equipment and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201208 |