Analysis method, system and equipment of three-dimensional point cloud data and storage medium

Abstract

The invention discloses a three-dimensional point cloud data analysis method, system and device and a storage medium. The method comprises the steps of obtaining sample three-dimensional point cloud data of a plurality of sample objects; establishing a first point cloud classification model through a preset nonparametric network algorithm according to the sample three-dimensional point cloud data; wherein the first point cloud classification model is a neural network model of a non-parametric network type; obtaining target three-dimensional point cloud data of a to-be-analyzed target object; and inputting the target three-dimensional point cloud data into the first point cloud classification model for reasoning analysis to obtain a first category prediction result corresponding to the target object. According to the method, the time consumption of the analysis process of the three-dimensional point cloud data can be effectively reduced, and the processing efficiency is improved; moreover, the method is suitable for the situation of storage noise or incomplete data in the three-dimensional point cloud data, and facilitates the improvement of the accuracy of analysis and processing. The method can be widely applied to the technical field of point cloud analysis.

Description

Technical Field

[0001] The present application relates to the technical field of point cloud analysis, and in particular to an analysis method, system, device and storage medium for three-dimensional point cloud data. Background Art

[0002] Currently, emerging application scenarios based on 3D (three-dimensional) vision are booming, and three-dimensional point clouds have attracted more and more extensive attention. Three-dimensional point clouds have a wide range of application fields, including but not limited to robotics, 3D graphics, autonomous driving, virtual reality (AR / VR / MR), etc. Among them, in the medical field, the application of three-dimensional point cloud technology is very extensive, covering multiple aspects from diagnosis, surgical planning to rehabilitation. In order to keep up with the growing application needs, the significance of developing analysis algorithms for three-dimensional point cloud data is significantly increasing.

[0003] In related technologies, when analyzing three-dimensional point cloud data, there is a need to determine the object category to which the three-dimensional point cloud data belongs, that is, to judge what kind of object the three-dimensional point cloud data is collected from. Currently, existing analysis methods generally use a neural network model to predict it. However, in practical applications, it is found that a large number of parameters need to be set when building the neural network model, and only after training the neural network model with a large amount of training data can a relatively satisfactory analysis effect be obtained. The overall implementation process takes a long time, resulting in low processing efficiency. Moreover, due to the fact that the three-dimensional point cloud data may contain some noise or incomplete data, existing technical solutions often have inaccurate problems during analysis.

[0004] In summary, the problems existing in related technologies need to be solved urgently. Summary of the Invention

[0005] An object of the present application is to solve at least to some extent one of the technical problems existing in related technologies.

[0006] To this end, an object of an embodiment of the present application is to provide an analysis method for three-dimensional point cloud data, which can effectively reduce the time-consuming of the analysis process of three-dimensional point cloud data, improve the processing efficiency, and is beneficial to improving the accuracy of analysis and processing.

[0007] To achieve the above technical object, the technical solutions adopted in the embodiments of the present application include:

[0008] On the one hand, an embodiment of the present application provides an analysis method for three-dimensional point cloud data, including:

[0009] Obtain the sample three-dimensional point cloud data of multiple sample objects;

[0010] Based on the sample three-dimensional point cloud data, a first point cloud classification model is established through a preset non-parametric network algorithm; wherein, the first point cloud classification model is a neural network model of the non-parametric network type;

[0011] Obtain the target three-dimensional point cloud data of the target object to be analyzed;

[0012] Input the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain the first category prediction result corresponding to the target object.

[0013] In addition, according to the analysis method of the three-dimensional point cloud data in the above embodiments of the present application, the following additional technical features may also be included:

[0014] Further, in an embodiment of the present application, the step of establishing a first point cloud classification model according to the sample three-dimensional point cloud data through a preset non-parametric network algorithm includes:

[0015] Obtain a preset non-parametric network algorithm; wherein, the non-parametric network algorithm includes any one of the K-nearest neighbor algorithm, kernel density estimation algorithm, Gaussian process algorithm, decision tree algorithm, and random forest algorithm;

[0016] Establish a first point cloud classification model according to the non-parametric network algorithm.

[0017] Further, in an embodiment of the present application, the non-parametric network algorithm is the K-nearest neighbor algorithm; the step of establishing a first point cloud classification model according to the non-parametric network algorithm includes:

[0018] Extract the first feature data of each sample three-dimensional point cloud data;

[0019] Select the first feature data corresponding to at least one sample three-dimensional point cloud data as the cluster center and establish several initial clustering clusters;

[0020] Calculate the first distance between the feature to be clustered and the cluster center; wherein, the feature to be clustered is the first feature data corresponding to the sample three-dimensional point cloud data that has not been added to the clustering cluster;

[0021] When the first distance between the feature to be clustered and the cluster center is less than or equal to the distance threshold, add the sample three-dimensional point cloud data corresponding to the feature to be clustered to the clustering cluster corresponding to the cluster center and update the cluster center coordinates of the clustering cluster; or, when the first distance between the feature to be clustered and any cluster center is greater than the distance threshold, use the feature to be clustered as a new cluster center and additionally establish an initial clustering cluster;

[0022] Determine the clustering categories of each of the clustering clusters according to the sample object corresponding to the sample three-dimensional point cloud data included in each of the clustering clusters, and obtain the first point cloud classification model.

[0023] Further, in an embodiment of the present application, the inputting the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain the first category prediction result corresponding to the target object includes:

[0024] Extract the second feature data of the target three-dimensional point cloud data;

[0025] Calculate the second distance between the second feature data and the centroid coordinates of each of the clustering clusters;

[0026] Divide the target three-dimensional point cloud data into the clustering cluster with the smallest corresponding second distance, and determine the first category prediction result corresponding to the target object according to the clustering category of the divided clustering cluster.

[0027] Further, in an embodiment of the present application, the method further includes:

[0028] Based on the first point cloud classification model as the basic architecture, establish a second point cloud classification model through a preset parameter network algorithm;

[0029] Perform parameter training on the second point cloud classification model to obtain a trained second point cloud classification model;

[0030] Input the target three-dimensional point cloud data into the trained second point cloud classification model for inference analysis to obtain the second category prediction result corresponding to the target object.

[0031] Further, in an embodiment of the present application, the performing parameter training on the second point cloud classification model to obtain a trained second point cloud classification model includes:

[0032] Obtain the category label corresponding to each of the sample objects;

[0033] According to the sample three-dimensional point cloud data, perform inference analysis through the second point cloud classification model to obtain the third category prediction result corresponding to the sample object;

[0034] Determine the training loss value according to the category label and the third category prediction result;

[0035] Update the parameters of the second point cloud classification model according to the loss value to obtain a trained second point cloud classification model.

[0036] Further, in an embodiment of the present application, the method further includes:

[0037] Obtain the trained third point cloud classification model; wherein, the third point cloud classification model is a neural network model of a parametric network type;

[0038] Input the target 3D point cloud data into the third point cloud classification model for inference analysis to obtain the fourth category prediction result corresponding to the target object;

[0039] Determine the comprehensive category prediction result corresponding to the target object according to the first category prediction result and the fourth category prediction result.

[0040] On the other hand, an analysis system for 3D point cloud data is also provided in an embodiment of the present application, including:

[0041] A first acquisition unit, configured to acquire sample 3D point cloud data of multiple sample objects;

[0042] A building unit, configured to build a first point cloud classification model according to the sample 3D point cloud data through a preset non-parametric network algorithm; wherein, the first point cloud classification model is a neural network model of a non-parametric network type;

[0043] A second acquisition unit, configured to acquire target 3D point cloud data of a target object to be analyzed;

[0044] An inference unit, configured to input the target 3D point cloud data into the first point cloud classification model for inference analysis to obtain the first category prediction result corresponding to the target object.

[0045] On the other hand, an embodiment of the present application provides a computer device, including:

[0046] At least one processor;

[0047] At least one memory, configured to store at least one program;

[0048] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned analysis method for 3D point cloud data.

[0049] On the other hand, an embodiment of the present application further provides a computer-readable storage medium, in which a program executable by a processor is stored, and the program executable by the processor is used to implement the above-mentioned analysis method for 3D point cloud data when executed by the processor.

[0050] The advantages and beneficial effects of the present application will be partially given in the following description, partially will become obvious from the following description, or will be understood through the practice of the present application:

[0051] A method for analyzing three-dimensional point cloud data disclosed in an embodiment of the present application includes obtaining sample three-dimensional point cloud data of multiple sample objects; establishing a first point cloud classification model according to the sample three-dimensional point cloud data through a preset non-parametric network algorithm, where the first point cloud classification model is a neural network model of non-parametric network type; obtaining target three-dimensional point cloud data of a target object to be analyzed; and inputting the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain a first category prediction result corresponding to the target object. The method in the present application can effectively reduce the time consumption of the analysis process of three-dimensional point cloud data and improve the processing efficiency. Moreover, this method is applicable to the situation where there is storage noise or incomplete data in the three-dimensional point cloud data, which is beneficial to improving the accuracy of analysis and processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following introduces the accompanying drawings related to the technical solutions in the embodiments of the present application or the prior art. It should be understood that the accompanying drawings below are only for conveniently and clearly presenting some embodiments of the technical solutions in the present invention. For those skilled in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 Schematic diagram of the implementation environment of a method for analyzing three-dimensional point cloud data provided in an embodiment of the present application;

[0054] Figure 2 Schematic diagram of the flow of a method for analyzing three-dimensional point cloud data provided in an embodiment of the present application;

[0055] Figure 3 Schematic diagram of the flow of establishing a first point cloud classification model provided in an embodiment of the present application;

[0056] Figure 4 Schematic diagram of the flow of performing inference analysis in a first point cloud classification model provided in an embodiment of the present application;

[0057] Figure 5 Schematic diagram of the flow of parameter training for a second point cloud classification model provided in an embodiment of the present application;

[0058] Figure 6 Schematic diagram of the flow of another method for analyzing three-dimensional point cloud data provided in an embodiment of the present application;

[0059] Figure 7 Schematic diagram of the structure of a system for analyzing three-dimensional point cloud data provided in an embodiment of the present application;

[0060] Figure 8It is a schematic structural diagram of a computer device provided in an embodiment of the present application. Detailed implementation manners

[0061] The present application will be further described below in conjunction with the accompanying drawings of the specification and specific embodiments. The described embodiments should not be construed as limiting the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present application.

[0062] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0064] First, several nouns involved in the present application are analyzed:

[0065] 1) Artificial intelligence (AI): It is a new technical science that studies, develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence; artificial intelligence is a branch of computer science. Artificial intelligence attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. The research in this field includes robots, speech recognition, image recognition, natural language processing and expert systems, etc. Artificial intelligence can simulate the information process of human consciousness and thinking. Artificial intelligence also uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, sense the environment, acquire knowledge and use knowledge to obtain the best results of theories, methods, technologies and application systems.

[0066] 2) Neural network model. A neural network model is a computational model that is inspired by the structure and function of the biological nervous system, especially the way neurons are connected in the brain. This model realizes functions such as learning, classifying and predicting data by simulating the interconnections between a large number of simple processing units (referred to as "artificial neurons" or simply "neurons"). The core idea of the neural network model is to learn the mapping relationship between input data and output results by adjusting the weights of these connections.

[0067] Currently, emerging application scenarios based on 3D (three-dimensional) vision are booming, and 3D point clouds have attracted more and more extensive attention. 3D point clouds have a wide range of application fields, including but not limited to robotics, 3D graphics, autonomous driving, virtual reality (AR / VR / MR), etc. Among them, in the medical field, the application of 3D point cloud technology is very extensive, covering multiple aspects from diagnosis, surgical planning to rehabilitation. In order to keep up with the growing application needs, the significance of developing analysis algorithms for 3D point cloud data is rising significantly.

[0068] Specifically, for example, in some applications, 3D point cloud data can be used to realize the reconstruction of medical images. By converting two-dimensional CT or MRI scan images into 3D point cloud data, doctors can more intuitively observe the shape, size and location of lesions. This is of great significance for the diagnosis of diseases such as tumors, fractures, and vascular malformations. In some applications, such as complex surgical operations (such as orthopedics, neurosurgery, cardiac surgery, etc.), doctors can use the patient's anatomical structure reconstructed from 3D point cloud data for virtual surgical simulation, plan the surgical path in advance, and reduce surgical risks.

[0069] In related technologies, when analyzing 3D point cloud data, there is a need to determine the object category to which the 3D point cloud data belongs, that is, to judge what kind of object the 3D point cloud data is collected from. For example, for 3D point cloud data in the medical field, it is hoped to analyze whether it belongs to data of lesions (such as tumors), or data of normal organs or tissues.

[0070] Currently, existing analysis methods generally use neural network models to make predictions. However, in practical applications, it is found that a large number of parameters need to be set when building a neural network model in this implementation method. After training the neural network model with a large amount of training data, a relatively satisfactory analysis effect can be obtained. The overall implementation process takes a long time, resulting in low processing efficiency. Moreover, due to the fact that 3D point cloud data may contain some noise or incomplete data, existing technical solutions often have inaccurate problems during analysis.

[0071] To solve the problems existing in the related art, the embodiments of the present application provide an analysis method, system, device and storage medium for three-dimensional point cloud data, which includes: obtaining the sample three-dimensional point cloud data of multiple sample objects; establishing a first point cloud classification model according to the sample three-dimensional point cloud data through a preset non-parametric network algorithm; wherein, the first point cloud classification model is a neural network model of non-parametric network type; obtaining the target three-dimensional point cloud data of the target object to be analyzed; inputting the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain the first category prediction result corresponding to the target object. The method in the embodiments of the present application can effectively reduce the time consumption of the analysis process of three-dimensional point cloud data and improve the processing efficiency; moreover, this method is applicable to the situation where there is noise or incomplete data stored in the three-dimensional point cloud data, which is beneficial to improving the accuracy of analysis and processing.

[0072] Applying the analysis method of three-dimensional point cloud data in the embodiments of the present application in the medical field can help doctors obtain more reliable and accurate information, and can significantly improve the efficiency and accuracy of diagnosis and treatment in aspects such as personalized medicine, precision treatment and rehabilitation.

[0073] Of course, the analysis method of three-dimensional point cloud data provided in the embodiments of the present application is not limited to being used in the medical field, and it can also be executed in other various application scenarios:

[0074] Exemplarily, in some embodiments, the method of the present application can be applied in the scenarios of autonomous driving and robot navigation. For example, in autonomous vehicles and robots, three-dimensional point cloud data generated by lidar (LiDAR) can be used to perceive the surrounding environment. By applying the method of the present application, the categories of various surrounding objects, such as obstacles, road boundaries, pedestrians, etc., can be effectively identified, which is convenient for path planning and obstacle avoidance.

[0075] Exemplarily, in some embodiments, the method of the present application can be applied in the scenarios of industrial inspection and manufacturing. For example, on the production line, three-dimensional scanners are used to obtain the three-dimensional point cloud data of products. By using the method of the present application, the types of products can be identified, which is convenient for distinguishing different product types and realizing the scheduling of products on the production line.

[0076] Exemplarily, in some embodiments, the method of the present application can be applied in the scenarios of cultural heritage protection. For example, three-dimensional scanning technology is used to obtain the high-precision three-dimensional point cloud data of cultural relics. By using the method of the present application to classify the cultural relics, and then combining the corresponding three-dimensional point cloud data for digital archiving, which is convenient for subsequent related research and display.

[0077] Of course, it should be noted that the above application scenarios only serve as examples and do not mean to limit the actual application of the methods in the embodiments of the present application. Those skilled in the art can understand that in different application scenarios, the methods provided in the embodiments of the present application can be used to perform specified tasks.

[0078] Please refer to Figure 1 , Figure 1 which is a schematic diagram of the implementation environment of the three-dimensional point cloud data analysis method provided by the embodiments of the present application. The main software and hardware entities of this implementation environment mainly include a user terminal 110 and a server 120, and the user terminal 110 is communicatively connected to the server 120. Among them, the three-dimensional point cloud data analysis method can be configured to be executed on the server 120 side and implemented based on the data interaction between the user terminal 110 and the server 120.

[0079] Exemplarily, for example, in some scenarios, the user terminal 110 can be used to collect relevant three-dimensional point cloud data, and the server 120 can configure a relevant neural network model to implement the analysis of three-dimensional point cloud data. When executing the analysis method provided in the embodiments of the present application, first, the user terminal 110 can collect a large number of sample three-dimensional point cloud data of multiple sample objects and send this data to the server 120. The server 120 obtains the sample three-dimensional point cloud data of the sample objects and, based on this sample three-dimensional point cloud data, establishes a first point cloud classification model through a preset non-parametric network algorithm. Then, when there is a specific analysis requirement, the user terminal 110 sends a request to the server 120. The server 120 obtains the target three-dimensional point cloud data of the target object to be analyzed, performs inference analysis using the first point cloud classification model, obtains the first category prediction result corresponding to the target object, and then can feedback the first category prediction result to the user terminal 110.

[0080] Specifically, the user terminal 110 in the present application can include, but is not limited to, any one or more of a smart watch, a smart phone, a computer, a personal digital assistant (PDA), a smart voice interaction device, a smart home appliance, or a vehicle-mounted terminal. The server 120 can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0081] A communication connection can be established between the user terminal 110 and the server 120 through a wireless network or a wired network. The wireless network or the wired network uses standard communication technologies and / or protocols. The network can be set as the Internet or any other network, such as any combination including but not limited to a Local Area Network (LAN), a Metropolitan Area Network (MAN), a Wide Area Network (WAN), a mobile, wired or wireless network, a private network or a virtual private network.

[0082] Of course, it can be understood that Figure 1 the implementation environment in Figure 1 is only an optional application scenario of the method for analyzing three-dimensional point cloud data provided in the embodiments of the present application. The actual application is not fixed to the

[0083] Next, in combination with Figure 1 the shown implementation environment, the method for analyzing three-dimensional point cloud data provided in the embodiments of the present application will be described in detail.

[0084] First, please refer to Figure 2 , Figure 2 which is a schematic flowchart of the method for analyzing three-dimensional point cloud data provided in the embodiments of the present application. Figure 2 The method for analyzing three-dimensional point cloud data shown in Figure 2 can be applied to relevant computer devices in the server 120, but is not limited to the above form.

[0085] Step 210: Obtain the sample three-dimensional point cloud data of multiple sample objects;

[0086] Step 220: According to the sample three-dimensional point cloud data, establish a first point cloud classification model through a preset non-parametric network algorithm; wherein, the first point cloud classification model is a neural network model of the non-parametric network type;

[0087] Step 230: Obtain the target three-dimensional point cloud data of the target object to be analyzed;

[0088] Step 240: Input the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain the first category prediction result corresponding to the target object.

[0089] In the embodiments of the present application, a method for analyzing three-dimensional point cloud data is provided. This method can effectively reduce the time-consuming of the analysis process of three-dimensional point cloud data and improve the processing efficiency; moreover, this method is applicable to the situation where there is storage noise or incomplete data in the three-dimensional point cloud data, which is beneficial to improving the accuracy of analysis and processing.

[0090] Specifically, when executing the analysis method of the three-dimensional point cloud data provided in the embodiment of the present application, first, multiple objects can be selected as samples. In the embodiment of the present application, these objects are recorded as sample objects. The specific type of the sample object can be determined according to the actual application scenario of the present application. For example, according to the type of the target object to be analyzed, an object that is relatively close to the target object can be selected as a sample object, so that the accuracy of the analysis of the target object can be effectively improved later. Exemplarily, for example, in some scenarios, the sample object and the target object can be related structures in the human body, such as teeth, jaws, spine, limb bones, etc. The three-dimensional point cloud data of these objects can be used to reconstruct and analyze bone morphology, and assist in the diagnosis of fractures, deformities or joint problems. In some scenarios, the sample object and the target object can be organs, such as the heart, liver, lungs, etc. The three-dimensional point cloud data of these objects can help doctors understand the shape, size and position of the organs more accurately, and assist in surgical planning or disease diagnosis.

[0091] In step 210, the number of sample objects determined is multiple, and the specific number is not limited in this application. For each sample object, its corresponding three-dimensional point cloud data can be obtained, and in the embodiment of this application, it is recorded as sample three-dimensional point cloud data. Here, three-dimensional point cloud data refers to a set of discrete points representing the surface of an object (i.e., the object in this application) in three-dimensional space. Each point usually contains three coordinate values ​​(x, y, z), indicating the position of the point in three-dimensional space. In addition, the point cloud data may also contain other attribute information, such as color (RGB), reflection intensity, normal vector, etc. It should be noted that the three-dimensional point cloud data is composed of a series of discrete points, rather than a continuous surface or volume, and the arrangement of the points in various places may be disordered and of different densities, which is not limited in this application.

[0092] In the embodiments of the present application, there are multiple ways to obtain sample 3D point cloud data of the sample object. For example, in some embodiments, a laser radar (LiDAR) device can be used to scan the sample object to obtain sample 3D point cloud data of the sample object; in some embodiments, multiple camera devices can be used to shoot the sample object from different angles, and the depth information can be calculated by parallax to generate sample 3D point cloud data of the sample object; in some embodiments, the depth information of the sample object can also be directly obtained by using a depth camera (such as Kinect, RealSense, etc.) to generate sample 3D point cloud data. In the embodiments of the present application, there is no limitation on the specific way to obtain the sample 3D point cloud data.

[0093] In step 220, based on the sample three-dimensional point cloud data, a point cloud classification model can be established through a preset non-parametric network algorithm. In the embodiment of the present application, it is denoted as the first point cloud classification model. Here, the first point cloud classification model is a neural network model of the non-parametric network type. A neural network model of the non-parametric network type refers to a class of machine learning models or networks that do not rely on a fixed set of parameters. Compared with traditional parametric models (such as deep neural networks, linear regression, etc.), non-parametric networks have greater flexibility in processing data because they do not need to pre-determine the complexity of the model or the number of parameters. Instead, they dynamically adjust the complexity of the model according to the scale and complexity of the data.

[0094] Therefore, compared with the situation in the prior art where a large number of parameters need to be set and the training takes a long time when building a neural network model, in the embodiment of the present application, by using a preset non-parametric network algorithm to build the first point cloud classification model of the non-parametric network type based on the sample three-dimensional point cloud data, the efficiency of model building and application can be effectively improved.

[0095] Exemplarily, in some embodiments, the preset non-parametric network algorithm in the present application may include any one of the k-nearest neighbor algorithm, kernel density estimation algorithm, Gaussian process algorithm, decision tree algorithm, and random forest algorithm. When building the first point cloud classification model, the non-parametric network algorithm can be obtained first, and then the first point cloud classification model can be established based on the sample three-dimensional point cloud data.

[0096] In step 230, after the first point cloud classification model is built, in actual application, the three-dimensional point cloud data of the target object to be analyzed can be obtained. In the embodiment of the present application, the three-dimensional point cloud data of the target object is denoted as the target three-dimensional point cloud data. The acquisition method of the target three-dimensional point cloud data is similar to that of the sample three-dimensional point cloud data in the previous steps and will not be elaborated here.

[0097] In step 240, for the obtained target three-dimensional point cloud data, it can be input into the first point cloud classification model for inference and analysis to obtain the class prediction result corresponding to the target object. In the embodiment of the present application, it is denoted as the first class prediction result. The first class prediction result can be used to characterize the class of the target object, and the class of the target object can be any one of multiple predetermined class types. It can be understood that in different scenarios, the class range of the target object can be flexibly set according to actual needs, and the present application does not limit this. Exemplarily, for example, in some scenarios, when analyzing whether a patient has a certain disease, by judging the class of the target object, it can be determined whether it belongs to the class of lesions (such as tumors, etc.), so as to help determine the patient's condition.

[0098] In the embodiment of the present application, a first point cloud classification model of a non-parametric network type is used to determine a first category prediction result corresponding to a target object, which can, to a certain extent, avoid the overfitting problem of a conventional neural network model. Because the complexity of the first point cloud classification model of the non-parametric network type is dynamically adjusted according to the data, it has a relatively better application effect for point cloud data of different scales and types. Moreover, the non-parametric network type can better resist the influence of noise and incomplete data, and has a wider applicability in practical applications.

[0099] It can be understood that the method for analyzing three-dimensional point cloud data provided in the embodiment of the present application includes: obtaining sample three-dimensional point cloud data of multiple sample objects; establishing a first point cloud classification model through a preset non-parametric network algorithm according to the sample three-dimensional point cloud data, where the first point cloud classification model is a neural network model of a non-parametric network type; obtaining target three-dimensional point cloud data of a target object to be analyzed; and inputting the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain a first category prediction result corresponding to the target object. The method in the embodiment of the present application can effectively reduce the time-consuming of the analysis process of three-dimensional point cloud data and improve the processing efficiency. Moreover, this method is applicable to the situation where there is noise or incomplete data stored in the three-dimensional point cloud data, which is beneficial to improving the accuracy of analysis and processing.

[0100] Specifically, in some embodiments, the non-parametric network algorithm is the K-nearest neighbor algorithm; referring to Figure 3 , according to the non-parametric network algorithm, establishing the first point cloud classification model includes:

[0101] Extracting first feature data of each of the sample three-dimensional point cloud data;

[0102] Selecting the first feature data corresponding to at least one of the sample three-dimensional point cloud data as cluster centers and establishing several initial clustering clusters;

[0103] Calculating a first distance between a feature to be clustered and the cluster centers, where the feature to be clustered is the first feature data corresponding to the sample three-dimensional point cloud data that has not been added to the clustering cluster;

[0104] When the first distance between the feature to be clustered and the cluster centers is less than or equal to the distance threshold, adding the sample three-dimensional point cloud data corresponding to the feature to be clustered to the clustering cluster corresponding to the cluster center and updating the cluster center coordinates of the clustering cluster; or, when the first distance between the feature to be clustered and any of the cluster centers is greater than the distance threshold, using the feature to be clustered as a new cluster center and additionally establishing an initial clustering cluster;

[0105] Determine the clustering categories of each of the clustering clusters according to the sample objects corresponding to the sample three-dimensional point cloud data included in each of the clustering clusters, so as to obtain the first point cloud classification model.

[0106] In some scenarios in the embodiments of the present application, the K-nearest neighbor algorithm can be used as a preset non-parametric network algorithm. At this time, the first point cloud classification model can be a clustering model, and when establishing the first point cloud classification model, a streaming clustering algorithm can be used.

[0107] Specifically, in the embodiments of the present application, first, the feature data corresponding to each sample three-dimensional point cloud data can be extracted and denoted as the first feature data. The first feature data can be extracted by using a relevant encoder or neural network model, and its data content can include values, vectors, matrices, etc., and the present application does not limit this.

[0108] After obtaining the first feature data corresponding to each sample three-dimensional point cloud data, at least one of the first feature data corresponding to the sample three-dimensional point cloud data can be selected as the initial cluster center to establish several initial clustering clusters. Then, for each sample three-dimensional point cloud data whose clustering category has not been determined, the first feature data corresponding to it is denoted as the feature to be clustered. According to the feature to be clustered and the position of the cluster center in the feature space, the distance between the feature to be clustered and each cluster center can be determined. In the embodiments of the present application, it is denoted as the first distance.

[0109] It can be understood that the first distance between the feature to be clustered and the initial cluster center can reflect the similarity between two sample three-dimensional point cloud data. The smaller the first distance, the closer the sample three-dimensional point cloud data to be clustered is to the clustering cluster. Therefore, in the embodiments of the present application, if the first distance between the feature to be clustered and a certain cluster center is less than or equal to the distance threshold, it means that the sample three-dimensional point cloud data corresponding to the feature to be clustered is similar enough to the sample three-dimensional point cloud data in the clustering cluster corresponding to the cluster center. Therefore, at this time, the sample three-dimensional point cloud data corresponding to the feature to be clustered can be added to the clustering cluster corresponding to the cluster center, and the cluster center coordinates of this clustering cluster are updated once. On the contrary, if the first distance between the feature to be clustered and all the already established initial cluster centers is greater than the distance threshold, it means that the sample three-dimensional point cloud data corresponding to the feature to be clustered is not similar to the sample three-dimensional point cloud data in each clustering cluster corresponding to the cluster centers. At this time, the sample three-dimensional point cloud data corresponding to the feature to be clustered needs to be divided into an additional category, that is, a new clustering cluster is established with the feature to be clustered as the cluster center. Repeat this process until all the sample three-dimensional point cloud data are divided into clustering clusters, and then the obtained clustering clusters can be used as the final clustering clusters.

[0110] In the embodiment of the present application, for each cluster obtained, the category of the sample object corresponding to the sample three-dimensional point cloud data contained therein can be determined. Since the sample objects in a cluster are similar, their categories are mostly the same. In the embodiment of the present application, the category of most sample objects in each cluster can be determined as the cluster category of the cluster, and the first point cloud classification model can be obtained. For the application in the present application, the cluster can be a category to which a group of objects belong in common. For example, the sample three-dimensional point cloud data corresponding to some categories of organs are currently obtained, and the sample three-dimensional point cloud data corresponding to organs belonging to the same category will be divided into the same cluster category.

[0111] Specifically, in some embodiments, referring to Figure 4 , the step of inputting the target three-dimensional point cloud data into the first point cloud classification model for reasoning and analysis to obtain a first category prediction result corresponding to the target object includes:

[0112] Extracting second feature data of the target three-dimensional point cloud data;

[0113] Calculating a second distance between the second feature data and the cluster center coordinates of each of the clustering clusters;

[0114] The target three-dimensional point cloud data is divided into the corresponding cluster with the smallest second distance, and the first category prediction result corresponding to the target object is determined according to the cluster category of the cluster to which the target object is divided.

[0115] In an embodiment of the present application, when determining the first category prediction result corresponding to the target object, reasoning analysis can be performed based on the first point cloud classification model. For example, when the first point cloud classification model adopts the clustering model in the above embodiment, the feature data of the target three-dimensional point cloud data can be extracted during reasoning analysis. In an embodiment of the present application, it is recorded as the second feature data. Then, the distance between the second feature data and the cluster center coordinates of each clustering cluster in the first point cloud classification model can be calculated. In an embodiment of the present application, it is recorded as the second distance. It can be understood that here, the method of extracting the second feature data is similar to the aforementioned first feature data, and the method of calculating the second distance is similar to the aforementioned first distance, which will not be elaborated in this application.

[0116] In the embodiment of the present application, after determining the second distance between the second feature data and the cluster center coordinates of each cluster, the target three-dimensional point cloud data can be divided into the corresponding cluster with the smallest second distance. Then, according to the cluster category of the cluster to which the target three-dimensional point cloud data is divided, the first category prediction result corresponding to the target object can be determined, that is, the cluster category of the cluster is determined as the category corresponding to the target object.

[0117] Specifically, in some embodiments, the method further includes:

[0118] Based on the first point cloud classification model as the basic architecture, a second point cloud classification model is established through a preset parameter network algorithm;

[0119] Perform parameter training on the second point cloud classification model to obtain a trained second point cloud classification model;

[0120] Input the target three-dimensional point cloud data into the trained second point cloud classification model for inference analysis to obtain a second category prediction result corresponding to the target object.

[0121] In the embodiments of the present application, in some scenarios, an extended application is also proposed. Using the previously proposed non-parameter network type first point cloud classification model as the basic architecture, another point cloud classification model of the parameter network type is constructed. In the embodiments of the present application, it can be denoted as the second point cloud classification model. Specifically, that is, combining the first point cloud classification model with some parameterized modules to construct the second point cloud classification model. The second point cloud classification model can have higher flexibility and scalability and can adapt to different application scenarios and requirements.

[0122] Exemplarily, in some embodiments, the present application can set some adaptive network layers on the basis of the first point cloud classification model. The adaptive network layer can be used to dynamically adjust the structure or parameters of the network according to the characteristics of the input data. For example, the adaptive network layer can adopt an attention mechanism, which can help locate which part of the data content the network needs to pay attention to. In some embodiments, the present application can add some meta-learning layers on the basis of the first point cloud classification model to enable the network to quickly adapt to new tasks or data sets. For example, by learning an initial parameter setting, the network can quickly converge when facing new tasks. This method combines the accuracy of the parameterized model and the adaptability of the non-parameter method, which can further improve the performance of the second point cloud classification model.

[0123] It should be noted that in the embodiments of the present application, for the established second point cloud classification model, it needs to be trained before being put into use. In actual use, the target three-dimensional point cloud data can be input into the trained second point cloud classification model for inference analysis to obtain the category prediction result corresponding to the target object. In the embodiments of the present application, the prediction result output by the second point cloud classification model is denoted as the second category prediction result.

[0124] Specifically, in some embodiments, referring to Figure 5 , the performing parameter training on the second point cloud classification model to obtain a trained second point cloud classification model includes:

[0125] Obtain the category labels corresponding to each of the sample objects;

[0126] Based on the sample three-dimensional point cloud data, perform inference analysis through the second point cloud classification model to obtain the third category prediction result corresponding to the sample object;

[0127] Determine the loss value of the training according to the category label and the third category prediction result;

[0128] Update the parameters of the second point cloud classification model according to the loss value to obtain a trained second point cloud classification model.

[0129] In the embodiment of the present application, when training the second point cloud classification model, the relevant data of the foregoing sample object can be used for training. Specifically, the category label corresponding to each sample object can be obtained, and this category label is used to characterize the true situation of the category of the sample object. Then, for each sample object, the second point cloud classification model can be used to perform inference analysis on its sample three-dimensional point cloud data to obtain the category prediction result corresponding to the sample object. In the embodiment of the present application, it is denoted as the third category prediction result. Exemplarily, for example, the category prediction result can be the true result indicating which organ category the sample object belongs to.

[0130] It can be understood that in the embodiment of the present application, if the prediction effect of the second point cloud classification model is good, then the third category prediction result corresponding to the sample object and the category label should be relatively close. Therefore, in the embodiment of the present application, the prediction accuracy of the second point cloud classification model can be determined based on the third category prediction result and the category label. Specifically, the deviation between the third category prediction result and the category label can be determined to obtain the loss value of the prediction of the second point cloud classification model. After obtaining the loss value, the prediction accuracy of the second point cloud classification model can be evaluated according to the magnitude of the loss value to perform backpropagation training on the second point cloud classification model and update its internal relevant parameters.

[0131] Specifically, for a machine learning model (or a deep learning model), the accuracy of its prediction can be measured by a loss function. The loss function is defined on a single training data and is used to measure the prediction error of a training data. Specifically, the loss value of a training data is determined by the label of the single training data and the prediction result of the model for the training data. During actual training, there are many training data in a training data set. Therefore, generally, a cost function is used to measure the overall error of the training data set. The cost function is defined on the entire training data set and is used to calculate the average value of the prediction errors of all training data, which can better measure the prediction effect of the model. For a general model, based on the aforementioned cost function, plus a regularization term that measures the complexity of the model, it can be used as the objective function for training. Based on this objective function, the loss value of the entire training data set can be obtained. There are many types of commonly used loss functions. For example, the 0-1 loss function, the square loss function, the absolute loss function, the logarithmic loss function, the cross-entropy loss function, etc. can all be used as the loss function of the machine learning model, which will not be elaborated one by one here. In the embodiments of the present application, any loss function can be selected to determine the loss value, so as to update the parameters of the second point cloud classification model.

[0132] In the embodiments of the present application, for the parameter update of the second point cloud classification model, a cyclic iteration method can be adopted. That is, after updating the parameters of the second point cloud classification model in one round, continue to use the second point cloud classification model with the updated parameters for prediction to determine a new loss value, and then update the parameters of the second point cloud classification model again. Repeat this process until the pre-set training end condition is met, and it can be considered that the training is completed, and the trained second point cloud classification model is obtained.

[0133] In the embodiments of the present application, the training end condition can be flexibly set according to requirements. For example, in some embodiments, the target number of rounds of training iteration can be set as the training end condition. When the number of update rounds of the parameters of the second point cloud classification model reaches the target number of rounds, it can be considered that the training is completed; in some embodiments, the difference threshold between the loss values obtained in two adjacent training processes can be set as the training end condition. After updating the parameters of the second point cloud classification model in one round, calculate the absolute value of the difference between the loss value obtained in this round of training process and the loss value obtained in the previous round of training process. If the absolute value of this difference is greater than the set difference threshold, continue the iterative training; if the absolute value of this difference is less than or equal to the set difference threshold, it can be considered that the training is completed. Of course, the above is only an exemplary introduction to some optional training end condition setting methods in the embodiments of the present application, and does not mean to limit the actual implementation situation.

[0134] Specifically, in some embodiments, referring toFigure 6 , the method further includes:

[0135] Obtain a trained third point cloud classification model; wherein, the third point cloud classification model is a neural network model of the parametric network type;

[0136] Input the target three-dimensional point cloud data into the third point cloud classification model for inference analysis to obtain a fourth category prediction result corresponding to the target object;

[0137] Determine a comprehensive category prediction result corresponding to the target object according to the first category prediction result and the fourth category prediction result.

[0138] In an embodiment of the present application, another method for analyzing three-dimensional point cloud data is further provided. This method can integrate a neural network model of the parametric network type and a neural network model of the non-parametric network type, analyze the target three-dimensional point cloud data through the two models, and then summarize the results of the two to obtain a comprehensive category prediction result.

[0139] Specifically, in an embodiment of the present application, a neural network model of the parametric network type can be additionally obtained. This neural network model can also be used to implement the analysis of point cloud data, denoted as the third point cloud classification model. For the third point cloud classification model, the target three-dimensional point cloud data can be input into the third point cloud classification model for inference analysis to obtain a category prediction result corresponding to the target object, which is denoted as the fourth category prediction result. Then, according to the foregoing first category prediction result and fourth category prediction result, a comprehensive category prediction result corresponding to the target object can be determined. For example, in some embodiments, the data forms of the first category prediction result and the fourth category prediction result can be vector forms, and each element in the vector represents the probability that the predicted target object belongs to a predetermined category. Through the method in the embodiment of the present application, the first category prediction result and the fourth category prediction result can be weighted and averaged, and the new vector obtained is output as the comprehensive category prediction result. The present application does not limit this.

[0140] Referring to Figure 7 , in an embodiment of the present application, an analysis system for three-dimensional point cloud data is further proposed, including:

[0141] A first acquisition unit 710, configured to acquire sample three-dimensional point cloud data of multiple sample objects;

[0142] A building unit 720, configured to build a first point cloud classification model according to the sample three-dimensional point cloud data through a preset non-parametric network algorithm; wherein, the first point cloud classification model is a neural network model of the non-parametric network type;

[0143] The second acquisition unit 730 is configured to acquire the target three-dimensional point cloud data of the target object to be analyzed;

[0144] The inference unit 740 is configured to input the target three-dimensional point cloud data into the first point cloud classification model for inference and analysis, so as to obtain the first category prediction result corresponding to the target object.

[0145] It can be understood that the content in the above embodiments of the three-dimensional point cloud data analysis method is applicable to the embodiments of this processing system. The functions specifically implemented in the embodiments of this processing system are the same as those in the above embodiments of the three-dimensional point cloud data analysis method, and the beneficial effects achieved are also the same as those in the above embodiments of the three-dimensional point cloud data analysis method.

[0146] Refer to Figure 8 , this application embodiment also discloses a computer device, including:

[0147] At least one processor 810;

[0148] At least one memory 820, configured to store at least one program;

[0149] When at least one program is executed by at least one processor 820, at least one processor 820 implements the above embodiments of the three-dimensional point cloud data analysis method.

[0150] It can be understood that the content in the above embodiments of the three-dimensional point cloud data analysis method is applicable to the embodiments of this computer device. The functions specifically implemented in the embodiments of this computer device are the same as those in the above embodiments of the three-dimensional point cloud data analysis method, and the beneficial effects achieved are also the same as those in the above embodiments of the three-dimensional point cloud data analysis method.

[0151] This application embodiment also discloses a computer-readable storage medium, in which a program executable by a processor is stored, and the program executable by the processor is used to implement the above embodiments of the three-dimensional point cloud data analysis method when executed by the processor.

[0152] It can be understood that the content in the above embodiments of the three-dimensional point cloud data analysis method is applicable to the embodiments of this computer-readable storage medium. The functions specifically implemented in the embodiments of this computer-readable storage medium are the same as those in the above embodiments of the three-dimensional point cloud data analysis method, and the beneficial effects achieved are also the same as those in the above embodiments of the three-dimensional point cloud data analysis method.

[0153] In some alternative embodiments, the functions / operations recited in the block diagrams may not occur in the order presented in the operational illustrations. For example, depending on the functions / operations involved, two blocks shown in succession may in fact be executed substantially simultaneously or the blocks may sometimes be executed in the reverse order. Further, the embodiments presented and described in the flowcharts of the present application are provided by way of example for the purpose of providing a more thorough understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and in which sub-operations described as part of a larger operation are performed independently.

[0154] In addition, although the present application has been described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It should also be understood that a detailed discussion of the actual implementation of each module is not necessary for an understanding of the present application. Rather, given the attributes, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the modules will be understood within the ordinary skill of an engineer. Accordingly, those skilled in the art can implement the present application as set forth in the claims without undue experimentation. It should also be understood that the specific concepts disclosed are illustrative only and are not intended to limit the scope of the present application, the scope of which is determined by the full scope of the appended claims and their equivalents.

[0155] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present application, in essence or the part that contributes to the prior art or a part of this technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc.

[0156] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definitional sequence of executable instructions for implementing logical functions, and can be embodied specifically in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with the instruction execution system, apparatus, or device.

[0157] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection portion having one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which a program can be printed, as the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other appropriate processing as necessary, and then stored in a computer memory.

[0158] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple 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 in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0159] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example", or "certain embodiments / examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0160] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present application. The scope of the present application is defined by the claims and their equivalents.

[0161] The above has specifically described the preferred embodiments of the present application, but the present application is not limited to the embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present application, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present application.

[0162] In the description of this specification, the description with reference to terms such as "one embodiment", "another embodiment", or "certain embodiments" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0163] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present application. The scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for analyzing three-dimensional point cloud data, characterized in that: include: Acquire sample three-dimensional point cloud data of multiple sample objects; According to the sample three-dimensional point cloud data, a first point cloud classification model is established by a preset non-parametric network algorithm; wherein the first point cloud classification model is a neural network model of a non-parametric network type; Acquire target three-dimensional point cloud data of the target object to be analyzed; The target three-dimensional point cloud data is input into the first point cloud classification model for reasoning and analysis to obtain a first category prediction result corresponding to the target object.

2. The method for analyzing three-dimensional point cloud data according to claim 1, characterized in that: The step of establishing a first point cloud classification model based on the sample three-dimensional point cloud data by using a preset non-parametric network algorithm includes: Obtain a preset non-parametric network algorithm; wherein the non-parametric network algorithm includes any one of a K-nearest neighbor algorithm, a kernel density estimation algorithm, a Gaussian process algorithm, a decision tree algorithm, and a random forest algorithm; According to the non-parametric network algorithm, a first point cloud classification model is established.

3. The method for analyzing three-dimensional point cloud data according to claim 2, characterized in that: The non-parametric network algorithm is a K-nearest neighbor algorithm; and establishing a first point cloud classification model according to the non-parametric network algorithm includes: Extracting first feature data of each of the sample three-dimensional point cloud data; Selecting the first feature data corresponding to at least one of the sample three-dimensional point cloud data as a cluster center, and establishing a plurality of initial clustering clusters; Calculating a first distance between the feature to be clustered and the cluster center; wherein the feature to be clustered is the first feature data corresponding to the sample three-dimensional point cloud data that is not added to the cluster; When the first distance between the feature to be clustered and the cluster center is less than or equal to the distance threshold, the sample three-dimensional point cloud data corresponding to the feature to be clustered is added to the cluster corresponding to the cluster center, and the cluster center coordinates of the cluster are updated; or, when the first distance between the feature to be clustered and any of the cluster centers is greater than the distance threshold, the feature to be clustered is used as a new cluster center, and an additional initial cluster is established; According to the sample objects corresponding to the sample three-dimensional point cloud data contained in each of the clusters, the clustering category of each of the clusters is determined to obtain the first point cloud classification model.

4. The method for analyzing three-dimensional point cloud data according to claim 3, characterized in that: The step of inputting the target three-dimensional point cloud data into the first point cloud classification model for reasoning and analysis to obtain a first category prediction result corresponding to the target object includes: Extracting second feature data of the target three-dimensional point cloud data; Calculating a second distance between the second feature data and the cluster center coordinates of each of the clustering clusters; The target three-dimensional point cloud data is divided into the corresponding cluster with the smallest second distance, and the first category prediction result corresponding to the target object is determined according to the cluster category of the cluster to which the target object is divided.

5. A method for analyzing three-dimensional point cloud data according to any one of claims 1 to 4, characterized in that: The method further comprises: Based on the first point cloud classification model as a basic framework, a second point cloud classification model is established through a preset parameter network algorithm; Performing parameter training on the second point cloud classification model to obtain a trained second point cloud classification model; The target three-dimensional point cloud data is input into the trained second point cloud classification model for reasoning and analysis to obtain a second category prediction result corresponding to the target object.

6. The method for analyzing three-dimensional point cloud data according to claim 5, characterized in that: The performing parameter training on the second point cloud classification model to obtain a trained second point cloud classification model includes: Obtaining the category label corresponding to each of the sample objects; According to the sample three-dimensional point cloud data, reasoning and analyzing are performed by the second point cloud classification model to obtain a third category prediction result corresponding to the sample object; Determining a training loss value according to the category label and the third category prediction result; According to the loss value, the parameters of the second point cloud classification model are updated to obtain a trained second point cloud classification model.

7. The method for analyzing three-dimensional point cloud data according to claim 1, characterized in that: The method further comprises: Acquire a trained third point cloud classification model; wherein the third point cloud classification model is a neural network model of a parameter network type; Inputting the target three-dimensional point cloud data into the third point cloud classification model for reasoning and analysis to obtain a fourth category prediction result corresponding to the target object; According to the first category prediction result and the fourth category prediction result, a comprehensive category prediction result corresponding to the target object is determined.

8. A three-dimensional point cloud data analysis system, characterized in that: include: A first acquisition unit, used to acquire sample three-dimensional point cloud data of a plurality of sample objects; An establishing unit, configured to establish a first point cloud classification model according to the sample three-dimensional point cloud data by a preset non-parametric network algorithm; wherein the first point cloud classification model is a neural network model of a non-parametric network type; A second acquisition unit is used to acquire target three-dimensional point cloud data of the target object to be analyzed; The inference unit is used to input the target three-dimensional point cloud data into the first point cloud classification model for inference analysis to obtain a first category prediction result corresponding to the target object.

9. A computer device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to implement the method according to any one of claims 1 to 7 when executed by the processor.

Images