Point cloud data feature processing system and method based on rotation

Abstract

The invention discloses a feature processing system and method for point cloud data based on rotation, and relates to the technical field of three-dimensional transformation processing of the point cloud data, and the method comprises the steps: obtaining target point cloud data, collecting and preprocessing the point cloud data of a target object through a three-dimensional scanning device, removing noise, and filling missing data to ensure the data quality; the invention relates to the technical field of three-dimensional transformation processing of point cloud data. According to the feature processing method and device for the point cloud data based on rotation and the terminal equipment, the rotation invariant features are extracted through the rotation mapping module, the problem of feature change of the point cloud data caused by rotation is effectively solved, the feature extraction method can reduce errors caused by view angle change, and the accuracy of feature extraction is improved. And the accuracy and robustness of point cloud data processing are improved, so that more reliable target identification is realized under various complex scenes and conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional transformation processing of point cloud data, and in particular to a feature processing system and method of point cloud data based on rotation. Background Art

[0002] In the field of 3D technology, point cloud data consists of a large number of disordered points, each of which contains its coordinate information in space. Due to the disorder and irregularity of point cloud data, traditional convolutional neural networks (CNNs) cannot be directly applied to the processing of point cloud data. To solve this problem, the research team at Stanford University proposed the PointNet model and introduced the T-Net rotation network into it to achieve normalized processing of point cloud data. The T-Net rotation network rotates the point cloud data so that the model can better learn the characteristics of the point cloud data.

[0003] However, in order to ensure the rotation invariance of point cloud data during 3D rotation, the prior art often provides a large amount of rotation enhancement data, that is, applies a spatial transformer network to normalize the input data before feature extraction. Although rotation enhancement data is effective in improving the rotation robustness of the machine model to some extent, it is obviously computationally expensive based on a large amount of rotation enhancement data, and the data enhancement is performed in a relatively fixed and standardized manner, so that the machine model is continuously trained in this fixed manner. The training method is not flexible enough and it is difficult to ensure rotation robustness. In this regard, we propose a feature processing method, device and terminal device based on rotation point cloud data to solve the above problems. Summary of the invention

[0004] In view of the above-mentioned existing problems, the present invention aims to solve several key problems in traditional technologies, including improving the accuracy of point cloud data acquisition and preprocessing, ensuring the integrity and quality of data through a data receiving unit and a preprocessing unit; solving the problem of rotation-invariant feature extraction, using a rotation mapping unit and a pooling processing unit to extract features that are not affected by rotation; improving the efficiency and accuracy of multi-dimensional feature processing, and realizing efficient and accurate feature representation through a clustering processing unit and a feature fusion unit; enhancing the accuracy of classification and recognition, and improving the performance of the recognition system with the help of a classifier training unit and a classification and recognition unit; and optimizing the management of data and results, and realizing effective management of point cloud data, features and classification results through a storage unit and a management unit, thereby comprehensively improving the overall performance and practicality of the system.

[0005] In order to solve the above technical problems, a feature processing system based on rotation point cloud data is proposed, which includes a data acquisition module, a feature extraction module, a multi-dimensional feature processing module, a classification and recognition module, and a storage and management module;

[0006] The output end of the data acquisition module is electrically connected to the input end of the feature extraction module;

[0007] The output end of the feature extraction module is electrically connected to the input end of the multi-dimensional feature processing module;

[0008] The output end of the multi-dimensional feature processing module is electrically connected to the input end of the classification recognition module;

[0009] The output end of the classification and identification module is electrically connected to the input end of the storage and management module.

[0010] As a preferred solution of a feature processing system based on rotation point cloud data described in the present invention, the data acquisition module includes a data receiving unit and a preprocessing unit, the data receiving unit acquires target point cloud data, and receives point cloud data from various sensors or devices, including lidar, camera, depth camera, wherein acquiring target point cloud data includes feature extraction and classification recognition of the three-dimensional coordinate information of the target object.

[0011] As a preferred solution of a feature processing system based on rotation point cloud data described in the present invention, wherein: the feature extraction module includes a rotation mapping unit and a pooling processing unit, which extracts rotation invariant features from the target point cloud data, and uses the rotation mapping unit to transform the target point cloud data from the original space to the feature space, thereby extracting the rotation invariant features and inputting them into the pooling processing unit for processing.

[0012] As a preferred solution of the feature processing system based on rotation point cloud data described in the present invention, wherein: the multidimensional feature processing module includes a clustering processing unit and a feature fusion unit, performs multidimensional feature processing on the extracted rotation invariant features, uses multiple clustering processing units to perform clustering processing on the rotation invariant features to obtain multidimensional feature representation, and performs feature fusion through the feature fusion unit;

[0013] The plurality of clustering processing units include a first clustering module, a second clustering module and a third clustering module, wherein:

[0014] The first clustering module performs preliminary clustering on the rotation invariant features with the maximum number of clusters;

[0015] The second clustering module further clusters the preliminary clustering results with a smaller number of clusters than the first clustering module;

[0016] The third clustering module performs final clustering on the further clustering results with a smaller number of clusters than that of the second clustering module.

[0017] The classification and recognition module includes a classifier training unit and a classification and recognition unit, which classifies and recognizes the rotation invariant features after multi-dimensional feature processing, and classifies the multi-dimensional features using the classifier training unit and the classification and recognition unit to obtain classification results.

[0018] As a preferred solution of the feature processing system based on rotation point cloud data described in the present invention, the storage and management module includes a storage unit and a management unit, which store and manage target point cloud data, extracted features and classification results.

[0019] Another object of the present invention is to provide a feature processing system for point cloud data based on rotation. The present invention aims to solve key technical problems in point cloud data processing, including achieving rotational invariance of point cloud data so that target objects can still be accurately identified and matched at different observation angles; improving the feature extraction and representation capabilities of point cloud data, generating rich and accurate multidimensional feature representations through multidimensional feature processing, thereby improving processing efficiency and accuracy; and optimizing the classification and recognition effect of point cloud data, using a neural network model to train rotational invariant features, and significantly improving the accuracy of classification and recognition.

[0020] As a preferred solution of the feature processing method based on rotation point cloud data described in the present invention, it is characterized by comprising:

[0021] Acquire target point cloud data, collect and pre-process the point cloud data of the target object through a three-dimensional scanning device, including removing noise and filling missing data; use a feature extraction module to transform the target point cloud data from the original space to the feature space, and extract rotation-invariant features; use multiple clustering processing units in the multidimensional feature processing module to perform multi-dimensional feature processing on the rotation-invariant features to generate a multi-dimensional feature representation, wherein the multiple clustering modules are connected in sequence according to the number of clusters from large to small, and each clustering processing unit performs clustering processing on the rotation-invariant features to generate clustering results at different levels; use a classification recognition module to classify the multi-dimensional feature representation to obtain a classification result.

[0022] As a preferred solution of the feature processing method based on rotation point cloud data described in the present invention, wherein: the extraction of rotation invariant features includes transforming the target point cloud data from the first dimensional space to the second dimensional space using a K nearest neighbor algorithm;

[0023] Inputting the transformed data into a first neural network model for training to obtain the rotation invariant feature;

[0024] The first neural network model is a convolutional neural network or a recurrent neural network.

[0025] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of the feature processing method based on rotation point cloud data are implemented.

[0026] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a feature processing method based on rotational point cloud data are implemented.

[0027] Beneficial effects of the present invention: The present invention extracts rotation-invariant features through a rotation mapping module, which effectively solves the problem of feature changes in point cloud data caused by rotation. This feature extraction method can reduce errors caused by changes in viewing angles, improve the accuracy and robustness of point cloud data processing, and thus achieve more reliable target recognition in various complex scenarios and conditions.

[0028] The present invention utilizes multiple clustering modules to perform multi-dimensional feature processing on rotation-invariant features to generate multi-dimensional feature representation. This multi-level, multi-dimensional feature processing method can capture more details and inherent laws of point cloud data, thereby enhancing the generalization ability of the deep learning model, which means that the model can better adapt to target objects of different types and forms, and improve the scope of application and performance of target recognition.

[0029] By utilizing advanced technologies such as neural network models and clustering algorithms, this patent can simplify the data processing process and reduce unnecessary human intervention and computing overhead, which not only reduces computing costs, but also improves the efficiency and speed of data processing, providing better support and guarantee for practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0031] Figure 1 A system solution module diagram of a feature processing system based on rotational point cloud data provided by an embodiment of the present invention.

[0032] Figure 2 A schematic diagram of a data acquisition module of a feature processing system based on rotational point cloud data provided by an embodiment of the present invention.

[0033] Figure 3 A schematic diagram of a feature extraction module of a feature processing system based on rotational point cloud data provided by an embodiment of the present invention.

[0034] Figure 4 A schematic diagram of a multi-dimensional feature processing module of a feature processing system based on rotational point cloud data provided by an embodiment of the present invention.

[0035] Figure 5 A schematic diagram of a classification and recognition module of a feature processing system based on rotational point cloud data provided by an embodiment of the present invention.

[0036] Figure 6 A schematic diagram of a storage and management module of a feature processing system for rotation-based point cloud data provided by an embodiment of the present invention.

[0037] Figure 7 An overall flow chart of a feature processing method of point cloud data based on rotation provided by an embodiment of the present invention.

[0038] In the figure: 01, data acquisition module; 02, feature extraction module; 03, multi-dimensional feature processing module; 04, classification and recognition module; 05, storage and management module; 011, data receiving unit; 012, preprocessing unit; 021, rotation mapping unit; 022, pooling processing unit; 031, clustering processing unit; 032, feature fusion unit; 041, classifier training unit; 042, classification and recognition unit; 051, storage unit; 052, management unit. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an embodiment that is mutually exclusive with other embodiments, either individually or selectively.

[0042] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0043] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0044] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0045] Example 1, reference Figure 1-Figure 6 , which is the first embodiment of the present invention, provides a feature processing system based on rotation point cloud data, including a data acquisition module 01, a feature extraction module 02, a multi-dimensional feature processing module 03, a classification and recognition module 04, and a storage and management module 05.

[0046] The output end of the data acquisition module 01 is electrically connected to the input end of the feature extraction module 02 .

[0047] The output end of the feature extraction module 02 is electrically connected to the input end of the multi-dimensional feature processing module 03 .

[0048] The output end of the multi-dimensional feature processing module 03 is electrically connected to the input end of the classification recognition module 04 .

[0049] The output end of the classification and identification module 04 is electrically connected to the input end of the storage and management module 05 .

[0050] The data acquisition module 01 is used to acquire the target point cloud data. This module can receive point cloud data from various sensors or devices, such as lidar, camera, depth camera. The acquired point cloud data contains the three-dimensional coordinate information of the target object, which can be used for subsequent feature extraction and classification recognition;

[0051] The feature extraction module 02 is used to extract rotation-invariant features from the target point cloud data. The module uses the rotation mapping module to transform the target point cloud data from the original space to the feature space, thereby extracting the rotation-invariant features;

[0052] The multi-dimensional feature processing module 03 is used to perform multi-dimensional feature processing on the extracted rotation invariant features. This module uses multiple clustering modules to perform clustering processing on the rotation invariant features to obtain multi-dimensional feature representation.

[0053] The classification and recognition module 04 is used to classify and recognize the rotation invariant features after multi-dimensional feature processing. This module uses the classifier module to classify the multi-dimensional features to obtain classification results. The storage and management module 05 is used to store and manage target point cloud data, extracted features and classification results.

[0054] Specifically, the data acquisition module 01 includes a data receiving unit 011 and a preprocessing unit 012. The data receiving unit 011 is used to receive raw point cloud data from a sensor or device. The unit can receive data by wired or wireless means and perform preliminary processing on the data, such as format conversion, data verification, etc. The preprocessing unit 012 is used to preprocess the received raw point cloud data to improve the accuracy of subsequent feature extraction. The preprocessing steps may include filtering, denoising, downsampling, etc.; for example, bilateral filtering, Gaussian filtering, etc. can be used to filter out noise points, and downsampling methods can be used to reduce the amount of point cloud data and improve processing efficiency.

[0055] Specifically, the feature extraction module 02 includes a rotation mapping unit 021 and a pooling processing unit 022. The rotation mapping unit 021 is used to extract rotation invariant features from the target point cloud data using the rotation mapping module. The unit first transforms the target point cloud data from the first dimensional space to the second dimensional space using the K nearest neighbor algorithm, and then inputs the transformed data into the first neural network model for training to obtain rotation invariant features.

[0056] During the rotation mapping process, the K nearest neighbor algorithm can be used to transform the target point cloud data from the original high-dimensional space to a low-dimensional space to reduce the computational complexity.

[0057] The transformed data is input into the first neural network model for training, which may be a convolutional neural network (CNN), a recurrent neural network (RNN) or other types of neural networks. Through training, the model can learn the rotation-invariant features of the target point cloud data. The pooling processing unit 022 is used to perform pooling on the extracted rotation-invariant features to reduce the dimension and number of features and improve the efficiency of subsequent processing. The pooling operation may be maximum pooling, average pooling or other types of pooling operations. Through pooling processing, a more compact and stable feature representation can be obtained.

[0058] Specifically, the multidimensional feature processing module 03 includes a clustering processing unit 031 and a feature fusion unit 032. The clustering processing unit 031 is used to perform clustering processing on the rotation invariant features using multiple clustering modules. The multiple clustering modules are sequentially connected in order of the number of clusters from large to small to achieve multi-level clustering of the rotation invariant features.

[0059] In a specific implementation, the rotation invariant features may be sequentially input into the first clustering module, the second clustering module, and the third clustering module for clustering processing. Each clustering module may be a K-means clustering, a hierarchical clustering, or other types of clustering algorithms. Through the clustering processing, clustering results at different levels may be obtained, thereby realizing a multi-dimensional feature representation of the rotation invariant features.

[0060] The feature fusion unit 032 is used to fuse the clustering results obtained by multiple clustering modules to obtain a more comprehensive and accurate feature representation. The fusion operation can be feature concatenation, feature weighting or other types of fusion operations. Through feature fusion, a multi-dimensional feature representation containing multiple levels of information can be obtained.

[0061] Specifically, the classification and recognition module 04 includes a classifier training unit 041 and a classification and recognition unit 042. The classifier training unit 041 is used to train the classifier module. During the training process, a large amount of labeled data can be used to train the classifier to improve its classification accuracy. The trained classifier can be used to classify and recognize multidimensional features. The classification and recognition unit 042 is used to classify and recognize multidimensional features using the trained classifier. The unit inputs the multidimensional features into the classifier to obtain a classification result. The classification result can be a category label, confidence or other form of classification information of the target object.

[0062] Specifically, the storage and management module 05 includes a storage unit 051 and a management unit 052. The storage unit 051 is used to store target point cloud data, extracted features, and classification results, etc. The storage unit 051 can be a storage device such as a hard disk, a solid-state drive (SSD), or a cloud storage. Through the storage unit 051, persistent storage and backup of data can be achieved. The management unit 052 is used to manage the stored data, including operations such as indexing, retrieving, and updating the data. The management unit 052 can provide a friendly user interface or API interface to facilitate users to operate and manage data.

[0063] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0064] Embodiment 2, the second embodiment of the present invention, is different from the first two embodiments in that:

[0065] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0066] The present invention also discloses a feature processing terminal device based on rotating point cloud data, including a memory and a processor, the memory is used to store programs and data; the processor is connected to the memory and is used to execute the program stored in the memory. In this embodiment, the terminal device can be a computer, a server, a mobile device or other types of terminal devices, and the memory is used to store target point cloud data, extracted features, and classification results. The processor is used to execute the computer program to implement the feature processing method based on rotating point cloud data.

[0067] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0068] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0069] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0070] Example 3, reference Figure 7 , which is the third embodiment of the present invention, provides a feature processing method of point cloud data based on rotation, comprising:

[0071] Step S1: Obtain target point cloud data, collect and pre-process the point cloud data of the target object through a three-dimensional scanning device, including removing noise and filling missing data to ensure data quality;

[0072] Step S2: using a rotation mapping module to transform the target point cloud data from the original space to the feature space to extract rotation invariant features, wherein the rotation mapping module includes: using a K nearest neighbor algorithm to transform the target point cloud data from the first dimensional space to the second dimensional space; inputting the transformed data into a first neural network model for training to obtain rotation invariant features;

[0073] Specifically, the first neural network model is a convolutional neural network, a recurrent neural network or any other type of neural network.

[0074] Step S3: using multiple clustering modules to perform multidimensional feature processing on the rotation invariant features to generate a multidimensional feature representation, wherein the multiple clustering modules are sequentially connected in descending order of the number of clusters, and each clustering module performs clustering processing on the rotation invariant features to generate clustering results at different levels;

[0075] Specifically, the multiple clustering modules include a first clustering module, a second clustering module and a third clustering module, wherein:

[0076] The first clustering module performs preliminary clustering of rotation invariant features with a larger number of clusters;

[0077] The second clustering module further clusters the preliminary clustering results with a smaller number of clusters than the first clustering module;

[0078] The third clustering module performs final clustering on the further clustering results with a smaller number of clusters than that of the second clustering module.

[0079] Step S4: using a classifier module to classify the multi-dimensional feature representation to obtain a classification result.

[0080] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A feature processing system based on rotation point cloud data, characterized in that: It includes data acquisition module, feature extraction module, multi-dimensional feature processing module, classification and recognition module, storage and management module; The output end of the data acquisition module is electrically connected to the input end of the feature extraction module; The output end of the feature extraction module is electrically connected to the input end of the multi-dimensional feature processing module; The output end of the multi-dimensional feature processing module is electrically connected to the input end of the classification and recognition module; The output end of the classification and identification module is electrically connected to the input end of the storage and management module.

2. A feature processing system based on rotation point cloud data as claimed in claim 1, characterized in that: The data acquisition module includes a data receiving unit and a preprocessing unit. The data receiving unit acquires target point cloud data and receives point cloud data from various sensors or devices, including lidar, camera, and depth camera. Acquiring target point cloud data includes feature extraction and classification recognition of the three-dimensional coordinate information of the target object.

3. A feature processing system based on rotation point cloud data as claimed in claim 2, characterized in that: The feature extraction module includes a rotation mapping unit and a pooling processing unit, which extracts rotation invariant features from target point cloud data. The rotation mapping unit is used to transform the target point cloud data from the original space to the feature space, thereby extracting the rotation invariant features and inputting them into the pooling processing unit for processing.

4. A feature processing system based on rotation point cloud data as claimed in claim 3, characterized in that: The multidimensional feature processing module includes a clustering processing unit and a feature fusion unit, performs multidimensional feature processing on the extracted rotation invariant features, uses multiple clustering processing units to perform clustering processing on the rotation invariant features to obtain multidimensional feature representation, and performs feature fusion through the feature fusion unit; The plurality of clustering processing units include a first clustering module, a second clustering module and a third clustering module, wherein: The first clustering module performs preliminary clustering on the rotation invariant features with the maximum number of clusters; The second clustering module further clusters the preliminary clustering results with a smaller number of clusters than the first clustering module; The third clustering module performs final clustering on the further clustering results with a smaller number of clusters than that of the second clustering module.

5. The feature processing method of point cloud data based on rotation according to claim 4, characterized in that: The classification and recognition module includes a classifier training unit and a classification and recognition unit, which classifies and recognizes the rotation invariant features after multi-dimensional feature processing, and classifies the multi-dimensional features using the classifier training unit and the classification and recognition unit to obtain classification results.

6. A feature processing method for point cloud data based on rotation as claimed in claim 5, characterized in that: The storage and management module includes a storage unit and a management unit, which store and manage target point cloud data, extracted features and classification results.

7. A method using a feature processing system based on rotation point cloud data as claimed in any one of claims 1 to 4, characterized in that: include, Obtain target point cloud data, collect and pre-process the point cloud data of the target object through 3D scanning equipment, including removing noise and filling missing data; The feature extraction module is used to transform the target point cloud data from the original space to the feature space and extract the rotation invariant features; Using a plurality of clustering processing units in a multidimensional feature processing module to perform multidimensional feature processing on the rotation invariant feature to generate a multidimensional feature representation, the plurality of clustering modules are sequentially connected in descending order of the number of clusters, and each clustering processing unit performs clustering processing on the rotation invariant feature to generate clustering results at different levels; The classification recognition module is used to classify the multi-dimensional feature representation to obtain the classification result.

8. The feature processing method of point cloud data based on rotation according to claim 7, characterized in that: The extracting of rotation invariant features includes transforming the target point cloud data from the first dimensional space to the second dimensional space using a K nearest neighbor algorithm; Inputting the transformed data into a first neural network model for training to obtain the rotation invariant feature; The first neural network model is a convolutional neural network or a recurrent neural network.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of a feature processing method based on rotation point cloud data according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a feature processing method based on rotation point cloud data described in any one of claims 1 to 7 are implemented.

Images