WO2022099528A1 - 点云法向量计算方法、装置、计算机设备和存储介质 - Google Patents

点云法向量计算方法、装置、计算机设备和存储介质 Download PDF

Info

Publication number
WO2022099528A1
WO2022099528A1 PCT/CN2020/128259 CN2020128259W WO2022099528A1 WO 2022099528 A1 WO2022099528 A1 WO 2022099528A1 CN 2020128259 W CN2020128259 W CN 2020128259W WO 2022099528 A1 WO2022099528 A1 WO 2022099528A1
Authority
WO
WIPO (PCT)
Prior art keywords
point
cloud data
point cloud
clustering
points
Prior art date
Application number
PCT/CN2020/128259
Other languages
English (en)
French (fr)
Inventor
吴伟
Original Assignee
深圳元戎启行科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳元戎启行科技有限公司 filed Critical 深圳元戎启行科技有限公司
Priority to PCT/CN2020/128259 priority Critical patent/WO2022099528A1/zh
Priority to CN202080092975.5A priority patent/CN114930402A/zh
Publication of WO2022099528A1 publication Critical patent/WO2022099528A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects

Definitions

  • the present application relates to a point cloud normal vector calculation method, apparatus, computer equipment and storage medium.
  • the point cloud normal vector can provide information related to the point cloud surface, and is widely used in 3D reconstruction, point cloud rendering, augmented reality, virtual reality, point cloud plane estimation and other fields.
  • the traditional way is to calculate the point cloud normal vector by calculating the covariance matrix eigenvalues and eigenvectors of a point and neighboring points.
  • the traditional method is easily interfered by adjacent objects in the process of selecting neighborhood points, resulting in low accuracy of the calculated point cloud normal vector.
  • a point cloud normal vector calculation method, apparatus, computer equipment and storage medium capable of improving the calculation accuracy of the point cloud normal vector.
  • a point cloud normal vector calculation method comprising:
  • a point cloud normal vector computing device comprising:
  • the acquisition module is used to acquire point cloud data
  • a segmentation module configured to invoke a pre-trained semantic segmentation model, input the point cloud data into the semantic segmentation model, and perform semantic segmentation on the point cloud data through the semantic segmentation model to obtain a semantic segmentation result;
  • a clustering module configured to cluster the point cloud data according to the semantic segmentation result to obtain a clustering result
  • a selection module for selecting the corresponding neighborhood points of each point in the point cloud data according to the clustering result
  • the calculation module is used to calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • a computer device comprising a memory and one or more processors, the memory having computer-readable instructions stored therein, the computer-readable instructions, when executed by the processor, cause the one or more processors to execute The following steps:
  • One or more non-volatile computer-readable storage media storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the following steps:
  • FIG. 1 is an application environment diagram of a method for calculating a point cloud normal vector in one or more embodiments.
  • FIG. 2 is a schematic flowchart of a method for calculating a point cloud normal vector in one or more embodiments.
  • FIG. 3 is a schematic flowchart of a step of clustering point cloud data according to a semantic segmentation result to obtain a clustering result in one or more embodiments.
  • FIG. 4 is a schematic flowchart of a step of selecting neighborhood points corresponding to each point in point cloud data according to a clustering result in one or more embodiments.
  • FIG. 5 is a schematic flowchart of a method for calculating a point cloud normal vector in another embodiment.
  • FIG. 6 is a block diagram of an apparatus for calculating point cloud normal vectors in one or more embodiments.
  • FIG. 7 is a block diagram of a computer device in one or more embodiments.
  • Figure 8 is a block diagram of a computer device in another embodiment.
  • the point cloud normal vector calculation method provided in this application can be applied to computer equipment, and the computer equipment can be a terminal or a server. It can be understood that the point cloud normal vector calculation method provided in this application can be applied to a terminal, a server, or a system including a terminal and a server, and is realized through interaction between the terminal and the server.
  • the point cloud normal vector calculation method provided in this application can be applied to the application environment shown in FIG. 1 .
  • the point cloud sensor 102 sends the collected point cloud data to the server 104 .
  • the point cloud sensor can be any one of the sensors used to collect point cloud data, such as lidar, laser scanner, camera, etc.
  • the server 104 invokes the pre-trained semantic segmentation model, inputs the point cloud data into the semantic segmentation model, and performs semantic segmentation on the point cloud data through the semantic segmentation model to obtain a semantic segmentation result. Therefore, the server 104 performs clustering on the point cloud data according to the semantic segmentation result to obtain a clustering result.
  • the computer device 104 selects a neighborhood point corresponding to each point in the point cloud data according to the clustering result, and then calculates a normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • a method for calculating a point cloud normal vector is provided, and the method is applied to a computer device as an example for description.
  • the computer device can be a terminal or a server, and includes the following steps:
  • Step 202 acquiring point cloud data.
  • the point cloud data may be data recorded by the point cloud sensor in the form of a point cloud by scanning the surrounding environment information.
  • Point cloud data is the point cloud data collected by point cloud sensors within the visible range. The visible range of different point cloud sensors can be different.
  • the point cloud data may specifically include three-dimensional coordinates (x, y, z) of each point, laser reflection intensity (Intensity), color information (RGB), and the like. Three-dimensional coordinates are used to represent the position information of the object surface in the surrounding environment.
  • the three-dimensional coordinates may be the coordinates of the point in the Cartesian coordinate system, and specifically include the horizontal, vertical, and vertical coordinates of the point in the Cartesian coordinate system.
  • the Cartesian coordinate system is a three-dimensional space coordinate system established with the point cloud sensor as the origin.
  • the three-dimensional space coordinate system includes a horizontal axis (x axis), a vertical axis (y axis) and a vertical axis (z axis).
  • the three-dimensional space coordinate system established with the point cloud sensor as the origin satisfies the right-hand rule.
  • the point cloud sensor obtains the corresponding point cloud data by scanning the current environment, and the vehicle-mounted sensor transmits the collected point cloud data to the computer equipment.
  • the onboard sensor could be a lidar.
  • Step 204 Invoke the pre-trained semantic segmentation model, input the point cloud data into the semantic segmentation model, and perform semantic segmentation on the point cloud data through the semantic segmentation model to obtain a semantic segmentation result.
  • a pre-trained semantic segmentation model is stored in the server, and the semantic segmentation model is obtained by training a large amount of sample data.
  • the server can train a deep learning model according to a large amount of manually labeled data carrying semantic category labels to obtain a semantic segmentation model, and training the deep learning model through data carrying semantic category labels can improve the Semantic segmentation accuracy.
  • the semantic segmentation model may be any one of the FCN (Fully Convolutional Networks, fully convolutional network) model, the conditional random field (CRF) model, pointnet, pointnet++ and other semantic segmentation models.
  • Semantic segmentation means that each point of the point cloud is divided into a corresponding category, that is, the semantic category corresponding to each point is given.
  • semantic categories may include: people, vehicles, roads, buildings, vegetation, guardrails, etc.
  • the server can perform semantic segmentation on the point cloud data through a pre-trained semantic segmentation model, and output the semantic segmentation result.
  • the semantic segmentation result includes the semantic category corresponding to each point in the point cloud data.
  • the computer device may further perform semantic segmentation on the point cloud data by using any semantic segmentation method such as edge-based, region-growth-based, attribute-based, and graph-based segmentation algorithms.
  • semantic segmentation method such as edge-based, region-growth-based, attribute-based, and graph-based segmentation algorithms.
  • the computer device may further preprocess the point cloud data before invoking the pre-trained semantic segmentation model.
  • Preprocessing can include clutter point removal and ground point filtering. Because there may be a large number of cluttered points in the point cloud data.
  • Computer equipment can remove a large number of cluttered points by performing pass-through filtering on the point cloud data. Further, the computer equipment can also perform ground point filtering processing on the point cloud data after removing the cluttered points.
  • the ground point filtering process refers to filtering out the ground points in the point cloud data, and the remaining points are non-ground points.
  • the computer equipment can perform ground segmentation on the point cloud data after removing the cluttered points, identify the ground points in the point cloud data after removing the cluttered points, and filter them to obtain non-ground point cloud data.
  • the computer device may firstly divide the point cloud area where the point cloud data from which the cluttered points are removed is located into a plurality of sub-areas.
  • the point cloud area refers to the three-dimensional data space where the point cloud data after removing the cluttered points is located.
  • the division method can be grid division of the point cloud data from which the cluttered points are removed, that is, the horizontal plane formed by the point cloud area in the x-axis direction and the y-axis direction is divided.
  • the grid division method can be equal division or random division. .
  • the point cloud area can be equally divided into 10*10 horizontal grids.
  • the computer device may use the least squares method to estimate the corresponding ground according to the preset plane equation, so as to obtain the corresponding ground of each sub-region.
  • the preset plane equation may be a ternary linear equation.
  • the ground corresponding to each sub-region is embodied in the form of a ternary linear equation.
  • the computer equipment traverses the coordinates of the points in the corresponding sub-regions in the equation corresponding to the ground, calculates the distance between each point and the corresponding ground, and determines the point as a ground point when the distance is less than a threshold. When the distance is greater than or equal to the threshold, the point is determined as a non-ground point. Threshold refers to the distance threshold used to judge whether the point is a ground point.
  • the computer equipment further filters the ground points to obtain non-ground point cloud data.
  • the computer equipment performs ground filtering on the point cloud data after removing the cluttered points, and can obtain effective point cloud data, that is, non-ground point cloud data. Furthermore, the computer equipment can perform semantic segmentation processing on the non-ground point cloud data.
  • Step 206 Cluster the point cloud data according to the semantic segmentation result to obtain a clustering result.
  • the computer equipment clusters the point cloud data according to the semantic category corresponding to each point in the semantic segmentation result to obtain a clustering result, which includes multiple The clustering category and the clustering category to which each point in the point cloud data belongs.
  • Clustering the point cloud data through the semantic segmentation results can divide the same points in the point cloud data into the same category, such as dividing a person and a vehicle in the point cloud data into a category, effectively avoiding the different The problem of clustering the points of semantic categories together.
  • the computer equipment can use a method based on connected domain analysis, K-means clustering method, Euclidean clustering, DBSCAN (Density-Based Spatial Clustering of Applications with Noise, density-based clustering with noise) Any one of the clustering methods, such as the density clustering algorithm, etc., can cluster the point cloud data.
  • Step 208 select neighbor points corresponding to each point in the point cloud data according to the clustering result.
  • the computer device can calculate the normal vector of the corresponding point by selecting the neighbor points of each point in the point cloud data, and using the selected neighbor points.
  • the computer equipment can select the neighborhood points through the clustering results.
  • the clustering result includes multiple clustering categories and the clustering category to which each point in the point cloud data belongs.
  • the computer device can determine the clustering category corresponding to the point according to the clustering result, and select points belonging to the same clustering category around the point to obtain the neighborhood corresponding to the point point.
  • the computer equipment selects the neighborhood points in the above manner, and then obtains the neighborhood points corresponding to all the points in the point cloud data.
  • the neighborhood points selected by the computer equipment can be a fixed number or points within a fixed sphere radius. Points in the same clustering category belong to the same semantic category. Therefore, the computer equipment selects the neighborhood points in the same clustering category, which can avoid the interference of adjacent objects, that is, avoid the problem that the neighborhood points of the A object are on the B object when the two objects A and B are close to each other. , thereby improving the accuracy of neighborhood point selection and improving the quality of neighborhood points.
  • Step 210 Calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • the normal vector corresponding to the point can be the point cloud normal vector, and the point cloud normal vector refers to the vector perpendicular to the surface plane of the point cloud.
  • the computer equipment can perform principal component analysis on the neighborhood points to calculate the normal vector corresponding to each point.
  • Principal component analysis refers to calculating the eigenvalues and eigenvectors of the covariance matrix of a point in the point cloud data and the corresponding neighboring points.
  • the computer device may calculate the normal vector corresponding to each point in the point cloud data according to the eigenvalues and eigenvectors of the covariance matrix.
  • calculating the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point includes: determining the covariance matrix corresponding to each point according to the neighborhood point corresponding to each point; calculating the covariance matrix corresponding to each point.
  • the eigenvalues, and the eigenvectors corresponding to the eigenvalues; according to the eigenvalues and the eigenvectors, the neighborhood points corresponding to each point are calculated to calculate the normal vectors corresponding to the corresponding points.
  • the computer device can determine the covariance matrix corresponding to the point according to the point and its corresponding neighborhood point, and perform eigenvalue decomposition processing on the covariance matrix to obtain the covariance matrix.
  • the eigenvalues of the matrix are calculated to obtain the eigenvectors corresponding to each eigenvalue, and the eigenvalues are arranged in a preset order.
  • the computer equipment determines the minimum eigenvalue of the covariance matrix, selects the eigenvector corresponding to the minimum eigenvalue, and normalizes the selected eigenvector to obtain the normal vector corresponding to each point in the point cloud data.
  • the normalization process refers to normalizing the selected feature vector into a unit vector.
  • the computer equipment obtains that the covariance matrix corresponding to a certain point is a 3 ⁇ 3 symmetric positive semi-definite matrix, then the computer equipment calculates and obtains three eigenvalues of the covariance matrix, and obtains the eigenvector corresponding to each eigenvalue, And arrange the three eigenvalues in descending order. Among the three eigenvalues, the eigenvector corresponding to the smallest eigenvalue is selected, and the selected eigenvector is normalized to obtain the normal vector corresponding to the point.
  • semantic segmentation is performed on the point cloud data through a semantic segmentation model to obtain a semantic segmentation result, so that the point cloud data is clustered according to the semantic segmentation result to obtain a clustering result.
  • the clustering results select the neighborhood points corresponding to each point in the point cloud data, and then calculate the normal vector corresponding to the corresponding point according to the neighborhood points corresponding to each point. Since the semantic segmentation result includes the semantic category corresponding to each point, the point cloud data is clustered according to the semantic segmentation result, which avoids the problem of clustering points of different semantic categories and improves the accuracy of clustering.
  • the point cloud data is clustered according to the semantic segmentation result, and the steps of obtaining the clustering result include:
  • Step 302 Perform connected domain detection on the point cloud data according to the semantic segmentation result to obtain multiple connected domains.
  • Step 304 Determine a plurality of clustering categories corresponding to the point cloud data and a clustering category corresponding to each point in the point cloud data according to the connected domain, and use the clustering category and the clustering category corresponding to each point in the point cloud data as the clustering result. .
  • the semantic segmentation result includes the semantic category corresponding to each point.
  • the computer equipment can perform connected domain detection on the point cloud data according to the semantic category corresponding to each point, and combine the points belonging to the same connected domain to obtain multiple connected domains. Connected domain detection can be done by spatially dividing the point cloud data, first selecting a point as the starting point, and extending outward from the starting point to obtain the relevant area, until there is no continuous point set in the neighborhood of the point, and finally Combine points belonging to the same connected domain together to get multiple connected domains.
  • a connected domain can correspond to one clustering category, so that the computer device can determine multiple clustering categories corresponding to the point cloud data and the clustering categories corresponding to each point in the point cloud data according to the connected domain, and will The clustering category corresponding to each point in the data generates the clustering result.
  • the connected domain detection is performed on the point cloud data according to the semantic segmentation result, and a plurality of cluster categories corresponding to the point cloud data and the clusters corresponding to each point in the point cloud data are determined according to the detected multiple connected domains category, which can combine points belonging to the same clustering category to achieve block processing.
  • the step of selecting a neighborhood point corresponding to each point in the point cloud data according to the clustering result includes:
  • Step 402 acquiring the current point in the point cloud data.
  • Step 404 in the clustering result, determine similar points belonging to the same clustering category as the current point.
  • Step 406 select a neighborhood point corresponding to the current point from the points of the same type.
  • Step 408 traverse all points in the point cloud data to obtain neighborhood points corresponding to each point in the point cloud data.
  • the computer device can directly select the neighborhood point corresponding to the point according to the clustering result.
  • the computer device can take any point in the point cloud data as the current point.
  • the clustering result includes the clustering category corresponding to each point in the point cloud data.
  • the computer device can determine the clustering category corresponding to the current point in the clustering result, and use all points in the clustering category corresponding to the current point as the same point. , so as to select neighbor points from the same kind of points.
  • the computer equipment traverses all the points in the point cloud data, and selects the neighborhood point corresponding to each point in the above-mentioned manner.
  • selecting a neighborhood point corresponding to a current point among similar points includes: selecting a preset distance among similar points according to a preset distance parameter. The number of points to get the neighborhood point corresponding to the current point.
  • a preset distance parameter and a preset number are pre-stored in the computer device.
  • the preset distance parameter can be the closest point, and the preset number can be 30.
  • the computer device selects the same point from the current point corresponding to the current point. The nearest 30 points.
  • selecting a neighborhood point corresponding to the current point from the same type of points includes: selecting a point from the same type of point that is within a preset radius of the sphere to obtain the neighborhood point corresponding to the current point.
  • a preset sphere radius is stored in the computer device, the computer device takes the current point as the center of the sphere, calculates the sphere space corresponding to the current point according to the preset sphere radius, and uses the same point in the sphere space as the neighborhood point corresponding to the current point.
  • the same type of points that belong to the same clustering category as the current point are determined in the clustering result, the neighboring points corresponding to the current point are selected from the same type of points, and all points in the point cloud data are traversed to obtain the point cloud data Neighborhood points corresponding to each point in . It is only necessary to select neighborhood points in the same clustering category, and the points of one object will not be used as the neighborhood points of another object, which effectively avoids the interference between adjacent objects and further improves the accuracy of neighborhood point selection. sex.
  • a method for calculating a point cloud normal vector which specifically includes the following steps:
  • Step 502 acquiring point cloud data.
  • Step 504 call the pre-trained semantic segmentation model, input the point cloud data into the semantic segmentation model, and perform semantic segmentation on the point cloud data through the semantic segmentation model to obtain a semantic segmentation result.
  • Step 506 Cluster the point cloud data according to the semantic segmentation result to obtain a clustering result.
  • Step 508 extract the clustering category in the clustering result.
  • Step 510 Count the number of points corresponding to the clustering categories according to the clustering result.
  • Step 512 remove all points corresponding to the cluster categories whose number of points is less than the threshold.
  • Step 514 selecting neighbor points corresponding to each point in the clustering result after removal processing.
  • Step 516 Calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • the computer equipment After acquiring the point cloud data, the computer equipment performs semantic segmentation on the point cloud data through a semantic segmentation model to obtain a semantic segmentation result, and then clusters the point cloud data according to the semantic segmentation result to obtain a clustering result.
  • the computer device can extract the clustering categories in the clustering result, and count the number of points corresponding to each clustering category.
  • a threshold is pre-stored in the computer device, and the threshold is used to determine whether the clustering category needs to be removed. The number of points corresponding to each clustering category is compared with the threshold. When the number of points is less than the threshold, the computer equipment will remove the points in the corresponding clustering category as noise points, and obtain the clustering result after removal.
  • the processed clustering result includes point cloud data with noise points removed.
  • the computer device When the number of points is greater than or equal to the threshold, the computer device retains the corresponding clustering category.
  • the computer equipment further performs neighborhood point selection for each point in the clustering result after removal. Specifically, for any point in the clustering result after the removal processing, the computer device may determine the clustering category corresponding to the point in the clustering result after the removal processing, and assign all the points in the clustering category corresponding to the point.
  • the points are regarded as the same kind of points, so that the neighboring points are selected from the same kind of points.
  • There are many ways for the computer equipment to select neighborhood points in the same kind of points and a preset number of points can be selected from the same kind of points according to the preset distance parameter to obtain the neighborhood points corresponding to the points.
  • the preset distance parameter may be the point with the closest distance, and the preset number may be 30, so that the computer device selects 30 points with the closest distance from the points of the same type corresponding to the point. It is also possible to select a point within the preset sphere radius from the same kind of points to obtain the corresponding neighborhood point of the point.
  • the computer device takes the point as the center of the sphere, calculates the sphere space corresponding to the point according to the preset sphere radius, and uses the same point in the sphere space as the neighborhood point corresponding to the point. It is only necessary to select neighborhood points in the same clustering category, and the points of one object will not be used as the neighborhood points of another object, which effectively avoids the interference between adjacent objects and further improves the accuracy of neighborhood point selection. sex.
  • the computer equipment further calculates the normal vector corresponding to the point according to the neighborhood point corresponding to each point, and obtains the normal vector corresponding to each point.
  • the computer device removes the points in the cluster category whose number of points is less than the threshold as noise points by counting the number of points of each cluster category in the clustering result. Retaining the cluster categories with the number of points greater than or equal to the threshold can remove the noise points in the point cloud data, avoid the problem of selecting noise points as neighborhood points, and improve the accuracy of neighborhood point selection. The calculation of the normal vector of the point is performed, thereby improving the calculation accuracy of the normal vector of the point.
  • a device for calculating point cloud normal vectors including: wherein:
  • the acquiring module 602 is used for acquiring point cloud data.
  • the segmentation module 604 is configured to invoke a pre-trained semantic segmentation model, input the point cloud data into the semantic segmentation model, and perform semantic segmentation on the point cloud data through the semantic segmentation model to obtain a semantic segmentation result.
  • the clustering module 606 is configured to cluster the point cloud data according to the semantic segmentation result to obtain a clustering result.
  • the selection module 608 is configured to select neighbor points corresponding to each point in the point cloud data according to the clustering result.
  • the calculation module 610 is configured to calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • the clustering module 606 is further configured to perform connected domain detection on the point cloud data according to the semantic segmentation result, to obtain multiple connected domains; and to determine multiple clustering categories and points corresponding to the point cloud data according to the connected domains
  • the clustering category corresponding to each point in the cloud data, and the clustering category and the clustering category corresponding to each point in the point cloud data are used as the clustering result.
  • the selection module 608 is further configured to obtain the current point in the point cloud data; according to the clustering result, determine the same type of point that belongs to the same clustering category as the current point; select the same type of point corresponding to the current point Neighborhood points; traverse all points in the point cloud data to obtain the corresponding neighborhood points of each point in the point cloud data.
  • the selection module 608 is further configured to select a preset number of points from the same type of points according to a preset distance parameter to obtain a neighborhood point corresponding to the current point.
  • the selection module 608 is further configured to select a point within the preset sphere radius from the current point among the same points to obtain a neighborhood point corresponding to the current point.
  • the calculation module 610 is further configured to determine the covariance matrix corresponding to each point according to the neighborhood point corresponding to each point; calculate the eigenvalue of the covariance matrix corresponding to each point, and the eigenvector corresponding to the eigenvalue ; Calculate the normal vector corresponding to the corresponding point according to the eigenvalue and the eigenvector to calculate the neighborhood point corresponding to each point.
  • the above-mentioned device further includes:
  • the removal module is used to extract the clustering category in the clustering result; count the number of points corresponding to the clustering category according to the clustering result; remove all points corresponding to the clustering category whose number of points is less than the threshold.
  • the selection module 608 is further configured to select neighborhood points corresponding to each point in the clustering result after removal processing.
  • the calculation module 610 is further configured to calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • Each module in the above-mentioned trajectory prediction apparatus can be implemented in whole or in part by software, hardware and combinations thereof.
  • the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
  • a computer device in one of the embodiments, the computer device may be a server, and its internal structure diagram may be as shown in FIG. 7 .
  • the computer device includes a processor, memory, a communication interface, and a database connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium, an internal memory.
  • the non-volatile storage medium stores an operating system, computer readable instructions and a database.
  • the internal memory provides an environment for the execution of the operating system and computer-readable instructions in the non-volatile storage medium.
  • the database of the computer device is used to store data of a method for calculating point cloud normal vectors.
  • the communication interface of the computer device is used to connect and communicate with an external terminal.
  • the computer readable instructions when executed by a processor, implement a method for computing point cloud normal vectors.
  • a computer device is provided, and the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 8 .
  • the computer equipment includes a processor, memory, a network interface, a display screen, and an input device connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium, an internal memory.
  • the non-volatile storage medium stores an operating system and computer readable instructions.
  • the internal memory provides an environment for the execution of the operating system and computer-readable instructions in the non-volatile storage medium.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer program implements a point cloud normal vector calculation method when executed by the processor.
  • the display screen of the computer equipment may be a liquid crystal display screen or an electronic ink display screen
  • the input device of the computer equipment may be a touch layer covered on the display screen, or a button, a trackball or a touchpad set on the shell of the computer equipment , or an external keyboard, trackpad, or mouse.
  • FIG. 7 or 8 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied.
  • a device may include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.
  • a computer device includes a memory and one or more processors, the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, causes the one or more processors to perform the following steps: obtaining Point cloud data; call the pre-trained semantic segmentation model, input the point cloud data into the semantic segmentation model, and perform semantic segmentation on the point cloud data through the semantic segmentation model to obtain the semantic segmentation results; cluster the point cloud data according to the semantic segmentation results , get the clustering result; select the neighborhood point corresponding to each point in the point cloud data according to the clustering result; calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • the processor further implements the following steps when executing the computer-readable instructions: performing connected domain detection on the point cloud data according to the semantic segmentation result to obtain multiple connected domains; determining multiple connected domains corresponding to the point cloud data according to the connected domains.
  • the clustering category and the clustering category corresponding to each point in the point cloud data, and the clustering category and the clustering category corresponding to each point in the point cloud data are used as the clustering result.
  • the processor when the processor executes the computer-readable instructions, it further implements the following steps: acquiring the current point in the point cloud data; determining, according to the clustering result, similar points that belong to the same clustering category as the current point; Select the neighborhood point corresponding to the current point; traverse all the points in the point cloud data, and obtain the neighborhood point corresponding to each point in the point cloud data.
  • the processor executes the computer-readable instructions, the following steps are further implemented: selecting a preset number of points from the same type of points according to a preset distance parameter to obtain a neighborhood point corresponding to the current point.
  • the processor executes the computer-readable instructions, the following steps are further implemented: selecting a point within a preset sphere radius from the current point among the same points, and obtaining a neighborhood point corresponding to the current point.
  • the processor further implements the following steps when executing the computer-readable instructions: determining a covariance matrix corresponding to each point according to the neighborhood points corresponding to each point; calculating the eigenvalues of the covariance matrix corresponding to each point, and The eigenvectors corresponding to the eigenvalues; according to the eigenvalues and eigenvectors, the neighborhood points corresponding to each point are calculated to calculate the normal vectors corresponding to the corresponding points.
  • the processor when the processor executes the computer-readable instructions, it further implements the following steps: extracting cluster categories in the clustering results; counting the number of points corresponding to the cluster categories according to the clustering results; All points corresponding to the class category are removed; the neighborhood points corresponding to each point in the clustering result after removal are selected; the normal vector corresponding to the corresponding point is calculated according to the neighborhood point corresponding to each point.
  • One or more non-volatile computer-readable storage media storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the following steps: acquiring a point cloud Data; call the pre-trained semantic segmentation model, input the point cloud data into the semantic segmentation model, and perform semantic segmentation on the point cloud data through the semantic segmentation model to obtain the semantic segmentation result; cluster the point cloud data according to the semantic segmentation result, and obtain Clustering result; select the neighborhood point corresponding to each point in the point cloud data according to the clustering result; calculate the normal vector corresponding to the corresponding point according to the neighborhood point corresponding to each point.
  • the following steps are further implemented: performing connected domain detection on the point cloud data according to the semantic segmentation result, to obtain a plurality of connected domains; Each clustering category and the clustering category corresponding to each point in the point cloud data are used as the clustering result.
  • the following steps are further implemented: obtaining the current point in the point cloud data; determining the same type of points that belong to the same clustering category as the current point according to the clustering result; Select the neighborhood point corresponding to the current point from the point; traverse all the points in the point cloud data, and obtain the neighborhood point corresponding to each point in the point cloud data.
  • the following steps are further implemented: selecting a preset number of points from the same type of points according to a preset distance parameter to obtain a neighborhood point corresponding to the current point.
  • the following steps are further implemented: selecting a point within a preset sphere radius from the current point among the same points to obtain a neighborhood point corresponding to the current point.
  • the following steps are further implemented: determining a covariance matrix corresponding to each point according to the neighborhood points corresponding to each point; calculating the eigenvalues of the covariance matrix corresponding to each point, and the eigenvector corresponding to the eigenvalue; according to the eigenvalue and the eigenvector, the neighborhood point corresponding to each point is calculated to calculate the normal vector corresponding to the corresponding point.
  • the following steps are further implemented: extracting the clustering categories in the clustering results; counting the number of points corresponding to the clustering categories according to the clustering results; All points corresponding to the clustering category are removed; the neighborhood points corresponding to each point in the clustering result after removal are selected; the normal vector corresponding to the corresponding point is calculated according to the neighborhood point corresponding to each point.
  • Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Road (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Abstract

一种点云法向量计算方法，包括：获取点云数据；调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果；根据语义分割结果对点云数据进行聚类，得到聚类结果；根据聚类结果选取点云数据中各点对应的邻域点；及根据各点对应的邻域点计算相应点对应的法向量。

Description

点云法向量计算方法、装置、计算机设备和存储介质 技术领域
本申请涉及一种点云法向量计算方法、装置、计算机设备和存储介质。
背景技术
随着传感器技术的进步，点云的获取越来越容易，而点云法向量作为最基本的点云特征，在诸多点云处理算法中起着至关重要的作用。点云法向量能够提供和点云表面相关的信息，被广泛应用于三维重建、点云渲染、增强现实、虚拟现实、点云平面估计等多种领域。
传统方式是通过计算一个点与邻域点的协方差矩阵特征值和特征向量来计算点云法向量。然而，传统方式在选取邻域点的过程中容易受到相邻物体的干扰，从而导致计算得到的点云法向量的准确性较低。
发明内容
根据本申请公开的各种实施例，提供一种能够提高点云法向量的计算准确性的点云法向量计算方法、装置、计算机设备和存储介质。
一种点云法向量计算方法，包括：
获取点云数据；
调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
根据各点对应的邻域点计算相应点对应的法向量。
一种点云法向量计算装置，包括：
获取模块，用于获取点云数据；
分割模块，用于调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
聚类模块，用于根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
选取模块，用于根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
计算模块，用于根据各点对应的邻域点计算相应点对应的法向量。
一种计算机设备，包括存储器和一个或多个处理器，所述存储器中储存有计算机可读指令，所述计算机可读指令被所述处理器执行时，使得所述一个或多个处理器执行以下步骤：
获取点云数据；
调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
根据各点对应的邻域点计算相应点对应的法向量。
一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行以下步骤：
获取点云数据；
调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
根据各点对应的邻域点计算相应点对应的法向量。
本申请的一个或多个实施例的细节在下面的附图和描述中提出。本申请的其它特征和优点将从说明书、附图以及权利要求书变得明显。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其它的附图。
图1为一个或多个实施例中点云法向量计算方法的应用环境图。
图2为一个或多个实施例中点云法向量计算方法的流程示意图。
图3为一个或多个实施例中根据语义分割结果对点云数据进行聚类，得到聚类结果步骤的流程示意图。
图4为一个或多个实施例中根据聚类结果选取点云数据中各点对应的邻域点步骤的流程示意图。
图5为另一个实施例中点云法向量计算方法的流程示意图。
图6为一个或多个实施例中点云法向量计算装置的框图。
图7为一个或多个实施例中计算机设备的框图。
图8为另一个实施例中计算机设备的框图。
具体实施方式
为了使本申请的技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请提供的点云法向量计算方法，可应用于计算机设备中，计算机设备可以为终端或服务器。可以理解的是，本申请提供的点云法向量计算方法可以应用于终端，也可以应用于服务器，还可以应用于包括终端和服务器的系统，并通过终端和服务器的交互实现。
本申请提供的点云法向量计算方法，可以应用于如图1所示的应用环境中。点云传感器102将采集到点云数据发送至服务器104。点云传感器可以是激光雷达、激光扫描仪、摄像头等用于采集点云数据的传感器中的任意一种。服务器104在获取到点云数据后，调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果。从而服务器104根据语义分割结果对点云数据进行聚类，得到聚类结果。计算机设备104根据聚类结果选取点云数据中各点对应的邻域点，进而根据各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，如图2所示，提供了一种点云法向量计算方法，以该方法应用于计算机设备为例进行说明，该计算机设备具体可以是终端或服务器，包括以下步骤：
步骤202，获取点云数据。
点云数据可以是点云传感器将扫描到的周围环境信息以点云形式记录的数据。点云数据是点云传感器在可视范围内采集到的点云数据。不同点云传感器的可视范围可以是不同的。点云数据具体可以包括各点的三维坐标(x，y，z)、激光反射强度(Intensity)、颜色信息(RGB)等。三维坐标用于表示周围环境中物体表面的位置信息。例如，三维坐标可以是点在笛卡尔坐标系中的坐标，具体包括点在笛卡尔坐标系中的横轴坐标、纵轴坐标和竖轴坐标。笛卡尔坐标系是以点云传感器为原点建立的三维空间坐标系，三维空间坐标系包括横轴(x轴)、纵轴(y轴)和竖轴(z轴)。以点云传感器为原点建立的三维空间坐标系满足右手定则。
具体的，点云传感器通过对当前环境进行扫描，得到相应的点云数据，车载传感器将采集到的点云数据传送至计算机设备。例如，车载传感器可以是激光雷达。
步骤204，调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果。
服务器中存储有预先训练的语义分割模型，语义分割模型是通过大量的样本数据训练得到的。在其中一个实施例中，服务器可以根据人工标注的大量携带有语义类别标签的数据对深度学习模型进行训练，得到语义分割模型，通过携带有语义类别标签的数据对深度学习模型进行训练，能够提高语义分割的准确性。例如，语义分割模型可以是FCN(Fully Convolutional Networks，全卷积网络)模型、条件随机场(conditional random field，简称CRF)模型、pointnet、pointnet++等语义分割模型中的任意一种。语义分割是指对点云的每个点都划分出对应的类别，即给出每个点对应的语义类别。例如，语义类别可以包括：人、车、道路、建筑、植被、护栏等。服务器在获取到点云数据后，可以通过预先训练的语义分割模型对点云数据进行语义分割，输出语义分割结果。语义分割结果中包括点云数据中各点对应的语义类别。
在其中一个实施例中，计算机设备还可以采用基于边缘的、基于区域增长的、基于属性的、基于图的分割算法等任意一种语义分割方法对点云数据进行语义分割。
在其中一个实施例中，计算机设备还可以在调用预先训练的语义分割模型之前，对点云数据进行预处理。预处理可包括杂乱点去除以及地面点过滤。由于点云数据中可能会存在大 量杂乱的点。计算机设备可以通过对点云数据进行直通滤波处理，从而去除大量杂乱的点。进一步的，计算机设备还可以对去除杂乱点后的点云数据进行地面点过滤处理。地面点过滤处理是指将点云数据中的地面点过滤掉，剩余的点则为非地面点。计算机设备可以通过对去除杂乱点后的点云数据进行地面分割，识别出去除杂乱点后的点云数据中的地面点，并进行过滤，得到非地面点云数据。具体的，计算机设备可以先将去除杂乱点后的点云数据所在的点云区域划分为多个子区域。点云区域是指去除杂乱点后的点云数据所在的三维数据空间。划分方式可以是对去除杂乱点的点云数据进行栅格划分，即对点云区域在x轴方向与y轴方向形成的水平面进行划分，栅格划分方式可以是均等划分，也可以是随机划分。例如，对于可视范围为100m的点云传感器，点云传感器扫描区域内的水平面大小为100m*100m，则可将点云区域均等划分成10*10个水平格子。针对划分得到的每个子区域，计算机设备可以根据预设平面方程采用最小二乘法估计相应的地面，从而得到每个子区域对应的地面。例如，预设平面方程可以是三元一次方程。每个子区域对应的地面是以三元一次方程的形式体现的。计算机设备在地面对应的方程中遍历输入相应子区域中的点坐标，计算各点与相应的地面之间的距离，当距离小于阈值时，则将该点确定为地面点。当距离大于或者等于阈值时，则将该点确定为非地面点。阈值是指用于判断该点是否为地面点的距离阈值。计算机设备进而将地面点进行过滤，从而得到非地面点云数据。计算机设备对去除杂乱点后的点云数据进行地面过滤处理，能够得到有效的点云数据，即非地面点云数据。进而计算机设备可以对非地面点云数据进行语义分割处理。
步骤206，根据语义分割结果对点云数据进行聚类，得到聚类结果。
由于语义分割结果中包括点云数据中各点对应的语义类别，计算机设备根据语义分割结果中各点对应的语义类别对点云数据进行聚类，得到聚类结果，聚类结果中包括多个聚类类别以及点云数据中各点所属的聚类类别。通过语义分割结果对点云数据进行聚类，能够将点云数据中相同的点划分为同一类别，如将点云数据中的某个人、某个车辆单独划分为一个类别，有效避免了将不同语义类别的点聚在一起的问题。
在其中一个实施例中，计算机设备可以采用基于连通域分析的方法、K均值聚类方法、欧几里德聚类、DBSCAN(Density-Based Spatial Clustering of Applications with Noise，具有噪声的基于密度的聚类方法)密度聚类算法等聚类方法中的任意一种对点云数据进行聚类。
步骤208，根据聚类结果选取点云数据中各点对应的邻域点。
计算机设备可以通过选取点云数据中各点的邻域点，并通过选取的邻域点来计算相应点的法向量。计算机设备可以通过聚类结果来进行邻域点的选取。具体的，聚类结果中包括多个聚类类别以及点云数据中各点所属的聚类类别。针对点云数据中的某个点，计算机设备可以根据聚类结果确定该点对应的聚类类别，并在该点的周围选取属于同一个聚类类别的点，从而得到该点对应的邻域点。计算机设备通过遍历点云数据中的所有点，采用上述方式进行邻域点选取，进而得到点云数据中所有点对应的邻域点。计算机设备选取的邻域点可以是固定数量的，也可以是固定球体半径内的点。同一个聚类类别中的点是属于同一个语义类别的。由此计算机设备在同一个聚类类别中选取邻域点，能够避免相邻物体的干扰，即避免两个物 体A和B挨得比较近时，A物体的邻域点在B物体上的问题，从而提高了邻域点选取的准确性，提高了邻域点的质量。
步骤210，根据各点对应的邻域点计算相应点对应的法向量。
点对应的法向量可以成为点云法向量，点云法向量是指垂直于该点云表平面的向量。计算机设备在选取各点对应的邻域点后，可以对邻域点进行主成分分析，来计算各点对应的法向量。主成分分析是指计算点云数据中一个点与对应的邻域点的协方差矩阵的特征值和特征向量。计算机设备可以根据协方差矩阵的特征值和特征向量来计算点云数据中各点对应的法向量。
在其中一个实施例中，根据各点对应的邻域点计算相应点对应的法向量包括：根据各点对应的邻域点确定各点对应的协方差矩阵；计算各点对应的协方差矩阵的特征值，以及特征值对应的特征向量；根据特征值以及特征向量计算各点对应的邻域点计算相应点对应的法向量。
具体的，针对点云数据中的每个点，计算机设备可以根据该点与该点对应的邻域点确定该点对应的协方差矩阵，对协方差矩阵进行特征值分解处理，可以得到协方差矩阵的特征值，从而计算得到每个特征值对应的特征向量，并将特征值按照预设顺序进行排列。计算机设备确定协方差矩阵的最小特征值，选取最小特征值对应的特征向量，对选取的特征向量进行归一化处理，得到得到点云数据中各点对应的法向量。归一化处理是指将选取的特征向量归一化为单位向量。例如，计算机设备得到某个点对应的协方差矩阵是一个3×3的对称半正定矩阵，则计算机设备计算得到协方差矩阵的特征值有3个，并得到每个特征值对应的特征向量，并将3个特征值按照降序顺序进行排列。在3个特征值中选取最小特征值对应的特征向量，对选取的特征向量进行归一化处理，进而得到该点对应的法向量。
在本实施例中，在获取点云数据后，通过语义分割模型对点云数据进行语义分割，得到语义分割结果，从而根据语义分割结果对点云数据进行聚类，得到聚类结果。根据聚类结果选取点云数据中各点对应的邻域点，进而根据各点对应的邻域点计算相应点对应的法向量。由于语义分割结果中包括各点对应的语义类别，根据语义分割结果对点云数据进行聚类，避免了将不同语义类别的点聚在一起的问题，提高了聚类的准确性。进一步根据聚类结果选取点云数据中各点对应的邻域点，能够确保邻域点是与各点在同一个聚类类别中的点，有效避免相邻物体的干扰，能够得到更为准确的邻域点，从而根据邻域点计算点的法向量，进而提高了点云法向量计算的准确性。另外，与基于深度学习的方法需要通过仿真的CAD模型来产生训练数据相比，是根据人工标注的大量携带有语义类别标签的数据进行训练得到的语义分割模型，是通过真实数据来训练模型，提高了语义分割的准确性。
在其中一个实施例中，如图3所示，根据语义分割结果对点云数据进行聚类，得到聚类结果的步骤包括：
步骤302，根据语义分割结果对点云数据进行连通域检测，得到多个连通域。
步骤304，根据连通域确定点云数据对应的多个聚类类别以及点云数据中各点对应的聚类类别，将聚类类别以及点云数据中各点对应的聚类类别作为聚类结果。
语义分割结果中包括各点对应的语义类别。计算机设备可以根据各点对应的语义类别对点云数据进行连通域检测，将属于同一连通域的点组合在一起，得到多个连通域。连通域检测的方式可以是对点云数据进行空间划分，先选定一个点作为起始点，由起始点向外延伸得到相关的区域，直至该点的邻域内不存在连续的点集为止，最后将属于同一个连通域的点组合在一起，得到多个连通域。一个连通域可以对应一个聚类类别，从而计算机设备可以根据连通域确定点云数据对应的多个聚类类别以及点云数据中各点对应的聚类类别，并将根据聚类类别以及点云数据中各点对应的聚类类别生成聚类结果。
在本实施例中，根据语义分割结果对点云数据进行连通域检测，并根据检测得到的多个连通域确定点云数据对应的多个聚类类别以及点云数据中各点对应的聚类类别，能够将属于同一个聚类类别的点组合在一起，实现分块处理。同时，有利于后续在同一个聚类类别中选取邻域点，提高邻域点选取的准确性。
在其中一个实施例中，如图4所示，根据聚类结果选取点云数据中各点对应的邻域点步骤包括：
步骤402，获取点云数据中的当前点。
步骤404，在聚类结果中确定与当前点属于相同聚类类别的同类点。
步骤406，在同类点中选取当前点对应的邻域点。
步骤408，遍历点云数据中的所有点，得到点云数据中各点对应的邻域点。
针对点云数据中的每个点，计算机设备可以直接根据聚类结果选取该点对应的邻域点。计算机设备可以将点云数据中的任意一个点作为作为当前点。聚类结果中包括点云数据中各点对应的聚类类别，计算机设备可以在聚类结果中确定当前点对应的聚类类别，并将当前点对应的聚类类别中的所有点作为同类点，从而在同类点中选取邻域点。计算机设备遍历点云数据中的所有点，采取上述方式选取每个点对应的邻域点。
计算机设备在同类点中选取邻域点的方式可以有多种，在其中一个实施例中，在同类点中选取当前点对应的邻域点包括：根据预设距离参数在同类点中选取预设数量的点，得到当前点对应的邻域点。计算机设备中预先存储有预设距离参数和预设数量，例如，预设距离参数可以是距离最近的点，预设数量可以是30，计算机设备从而在当前点对应的同类点中选取与当前点距离最近的30个点。
在其中一个实施例中，在同类点中选取当前点对应的邻域点包括：在同类点中选取与当前点在预设球体半径内的点，得到当前点对应的邻域点。计算机设备中存储有预设球体半径，计算机设备以当前点为球心，根据预设球体半径计算得到当前点对应的球体空间，将球体空间内的同类点作为当前点对应的邻域点。
在本实施例中，在聚类结果中确定与当前点属于相同聚类类别的同类点，在同类点中选取当前点对应的邻域点，遍历点云数据中的所有点，得到点云数据中各点对应的邻域点。只需要在同一个聚类类别中选取邻域点，不会将一个物体的点作为另外一个物体的邻域点，有效避免了相邻物体之间的干扰，进一步提高了邻域点选取的准确性。
在另一个实施例中，如图5所示，提供了一种点云法向量计算方法，具体包括以下步骤：
步骤502，获取点云数据。
步骤504，调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果。
步骤506，根据语义分割结果对点云数据进行聚类，得到聚类结果。
步骤508，提取聚类结果中的聚类类别。
步骤510，根据聚类结果统计聚类类别对应的点数量。
步骤512，将点数量小于阈值的聚类类别对应的所有点进行去除处理。
步骤514，选取去除处理后的聚类结果中各点对应的邻域点。
步骤516，根据各点对应的邻域点计算相应点对应的法向量。
计算机设备在获取到点云数据后，通过语义分割模型对点云数据进行语义分割，得到语义分割结果，从而根据语义分割结果对点云数据进行聚类，得到聚类结果。计算机设备可以提取聚类结果中的聚类类别，并统计每个聚类类别对应的点数量。计算机设备中预先存储有阈值，该阈值用于判断聚类类别是否需要去除。将每个聚类类别对应的点数量与阈值进行比较，当点数量小于阈值时，计算机设备将相应的聚类类别中的点作为噪声点，进行去除，得到去除处理后的聚类结果，去除处理后的聚类结果中包括去除噪声点的点云数据。当点数量大于或者等于阈值时，计算机设备则保留相应的聚类类别。计算机设备进而对去除处理后的聚类结果中的各点进行邻域点选取。具体的，针对去除处理后的聚类结果中的任意一个点，计算机设备可以在去除处理后的聚类结果中确定该点对应的聚类类别，并将该点对应的聚类类别中的所有点作为同类点，从而在同类点中选取邻域点。计算机设备在同类点中选取邻域点的方式可以有多种，可以根据预设距离参数在同类点中选取预设数量的点，得到该点对应的邻域点。例如，预设距离参数可以是距离最近的点，预设数量可以是30，计算机设备从而在该点对应的同类点中选取与该点距离最近的30个点。也可以在同类点中选取与该点在预设球体半径内的点，得到该点对应的邻域点。计算机设备以该点为球心，根据预设球体半径计算得到该点对应的球体空间，将球体空间内的同类点作为该点对应的邻域点。只需要在同一个聚类类别中选取邻域点，不会将一个物体的点作为另外一个物体的邻域点，有效避免了相邻物体之间的干扰，进一步提高了邻域点选取的准确性。计算机设备进而根据各点对应的邻域点计算该点对应的法向量，得到各点对应的法向量。
在本实施例中，计算机设备通过统计聚类结果中每个聚类类别的点数量，将点数量小于阈值的聚类类别中的点作为噪声点，进行去除。保留点数量大于或者等于阈值的聚类类别，能够将点云数据中的噪声点去除，避免了将噪音点选取为邻域点的问题，提高了邻域点选取的准确性，根据邻域点进行点的法向量的计算，从而提高了点的法向量的计算准确性。
在其中一个实施例中，如图6所示，提供了一种点云法向量计算装置，包括：其中：
获取模块602，用于获取点云数据。
分割模块604，用于调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果。
聚类模块606，用于根据语义分割结果对点云数据进行聚类，得到聚类结果。
选取模块608，用于根据聚类结果选取点云数据中各点对应的邻域点。
计算模块610，用于根据各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，聚类模块606，还用于根据语义分割结果对点云数据进行连通域检测，得到多个连通域；根据连通域确定点云数据对应的多个聚类类别以及点云数据中各点对应的聚类类别，将聚类类别以及点云数据中各点对应的聚类类别作为聚类结果。
在其中一个实施例中，选取模块608，还用于获取点云数据中的当前点；根据聚类结果中确定与当前点属于相同聚类类别的同类点；在同类点中选取当前点对应的邻域点；遍历点云数据中的所有点，得到点云数据中各点对应的邻域点。
在其中一个实施例中，选取模块608，还用于根据预设距离参数在同类点中选取预设数量的点，得到当前点对应的邻域点。
在其中一个实施例中，选取模块608，还用于在同类点中选取与当前点在预设球体半径内的点，得到当前点对应的邻域点。
在其中一个实施例中，计算模块610，还用于根据各点对应的邻域点确定各点对应的协方差矩阵；计算各点对应的协方差矩阵的特征值，以及特征值对应的特征向量；根据特征值以及特征向量计算各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，上述装置还包括：
去除模块，用于提取聚类结果中的聚类类别；根据聚类结果统计聚类类别对应的点数量；将点数量小于阈值的聚类类别对应的所有点进行去除处理。
选取模块608，还用于选取去除处理后的聚类结果中各点对应的邻域点。
计算模块610，还用于根据各点对应的邻域点计算相应点对应的法向量。
关于点云法向量计算装置的具体限定可以参见上文中对于点云法向量计算方法的限定，在此不再赘述。上述轨迹预测装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在其中一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图7所示。该计算机设备包括通过系统总线连接的处理器、存储器、通信接口和数据库。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机可读指令和数据库。该内存储器为非易失性存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的数据库用于存储一种点云法向量计算方法的数据。该计算机设备的通信接口用于与外部的终端连接通信。该计算机可读指令被处理器执行时以实现一种点云法向量计算方法。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是终端，其内部结构图可以如图8所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口、显示屏和输入装置。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统和计算机可读指 令。该内存储器为非易失性存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种点云法向量计算方法。该计算机设备的显示屏可以是液晶显示屏或者电子墨水显示屏，该计算机设备的输入装置可以是显示屏上覆盖的触摸层，也可以是计算机设备外壳上设置的按键、轨迹球或触控板，还可以是外接的键盘、触控板或鼠标等。
本领域技术人员可以理解，图7或8中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
一种计算机设备，包括存储器及一个或多个处理器，存储器中储存有计算机可读指令，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行以下步骤：获取点云数据；调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果；根据语义分割结果对点云数据进行聚类，得到聚类结果；根据聚类结果选取点云数据中各点对应的邻域点；根据各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，处理器执行计算机可读指令时还实现以下步骤：根据语义分割结果对点云数据进行连通域检测，得到多个连通域；根据连通域确定点云数据对应的多个聚类类别以及点云数据中各点对应的聚类类别，将聚类类别以及点云数据中各点对应的聚类类别作为聚类结果。
在其中一个实施例中，处理器执行计算机可读指令时还实现以下步骤：获取点云数据中的当前点；根据聚类结果中确定与当前点属于相同聚类类别的同类点；在同类点中选取当前点对应的邻域点；遍历点云数据中的所有点，得到点云数据中各点对应的邻域点。
在其中一个实施例中，处理器执行计算机可读指令时还实现以下步骤：根据预设距离参数在同类点中选取预设数量的点，得到当前点对应的邻域点。
在其中一个实施例中，处理器执行计算机可读指令时还实现以下步骤：在同类点中选取与当前点在预设球体半径内的点，得到当前点对应的邻域点。
在其中一个实施例中，处理器执行计算机可读指令时还实现以下步骤：根据各点对应的邻域点确定各点对应的协方差矩阵；计算各点对应的协方差矩阵的特征值，以及特征值对应的特征向量；根据特征值以及特征向量计算各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，处理器执行计算机可读指令时还实现以下步骤：提取聚类结果中的聚类类别；根据聚类结果统计聚类类别对应的点数量；将点数量小于阈值的聚类类别对应的所有点进行去除处理；选取去除处理后的聚类结果中各点对应的邻域点；根据各点对应的邻域点计算相应点对应的法向量。
一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，计算机可读指令被一个或多个处理器执行时，使得一个或多个处理器执行执行以下步骤：获取点云数据；调用预先训练的语义分割模型，将点云数据输入至语义分割模型，通过语义分割模型对点云数据进行语义分割，得到语义分割结果；根据语义分割结果对点云数据进行聚类，得到聚类结果； 根据聚类结果选取点云数据中各点对应的邻域点；根据各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，计算机可读指令被处理器执行时还实现以下步骤：根据语义分割结果对点云数据进行连通域检测，得到多个连通域；根据连通域确定点云数据对应的多个聚类类别以及点云数据中各点对应的聚类类别，将聚类类别以及点云数据中各点对应的聚类类别作为聚类结果。
在其中一个实施例中，计算机可读指令被处理器执行时还实现以下步骤：获取点云数据中的当前点；根据聚类结果中确定与当前点属于相同聚类类别的同类点；在同类点中选取当前点对应的邻域点；遍历点云数据中的所有点，得到点云数据中各点对应的邻域点。
在其中一个实施例中，计算机可读指令被处理器执行时还实现以下步骤：根据预设距离参数在同类点中选取预设数量的点，得到当前点对应的邻域点。
在其中一个实施例中，计算机可读指令被处理器执行时还实现以下步骤：在同类点中选取与当前点在预设球体半径内的点，得到当前点对应的邻域点。
在其中一个实施例中，计算机可读指令被处理器执行时还实现以下步骤：根据各点对应的邻域点确定各点对应的协方差矩阵；计算各点对应的协方差矩阵的特征值，以及特征值对应的特征向量；根据特征值以及特征向量计算各点对应的邻域点计算相应点对应的法向量。
在其中一个实施例中，计算机可读指令被处理器执行时还实现以下步骤：提取聚类结果中的聚类类别；根据聚类结果统计聚类类别对应的点数量；将点数量小于阈值的聚类类别对应的所有点进行去除处理；选取去除处理后的聚类结果中各点对应的邻域点；根据各点对应的邻域点计算相应点对应的法向量。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机可读指令来指令相关的硬件来完成，所述的计算机可读指令可存储于一非易失性计算机可读取存储介质中，该计算机可读指令在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不 脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种点云法向量计算方法，包括：
    获取点云数据；
    调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
    根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
    根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
    根据各点对应的邻域点计算相应点对应的法向量。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果包括：
    根据所述语义分割结果对所述点云数据进行连通域检测，得到多个连通域；及
    根据所述连通域确定所述点云数据对应的多个聚类类别以及所述点云数据中各点对应的聚类类别，将所述聚类类别以及所述点云数据中各点对应的聚类类别作为聚类结果。
  3. 根据权利要求1所述的方法，其特征在于，所述根据所述聚类结果选取所述点云数据中各点对应的邻域点包括：
    获取所述点云数据中的当前点；
    根据所述聚类结果中确定与所述当前点属于相同聚类类别的同类点；
    在所述同类点中选取所述当前点对应的邻域点；及
    遍历所述点云数据中的所有点，得到所述点云数据中各点对应的邻域点。
  4. 根据权利要求3所述的方法，其特征在于，在所述同类点中选取所述当前点对应的邻域点包括：
    根据预设距离参数在所述同类点中选取预设数量的点，得到所述当前点对应的邻域点。
  5. 根据权利要求3所述的方法，其特征在于，在所述同类点中选取所述当前点对应的邻域点包括：
    在所述同类点中选取与所述当前点在预设球体半径内的点，得到所述当前点对应的邻域点。
  6. 根据权利要求1所述的方法，其特征在于，所述根据各点对应的邻域点计算相应点对应的法向量包括：
    根据各点对应的邻域点确定各点对应的协方差矩阵；
    计算各点对应的协方差矩阵的特征值，以及所述特征值对应的特征向量；及
    根据所述特征值以及所述特征向量计算各点对应的邻域点计算相应点对应的法向量。
  7. 根据权利要求1至6任意一项所述的方法，其特征在于，所述方法还包括：
    提取所述聚类结果中的聚类类别；
    根据所述聚类结果统计所述聚类类别对应的点数量；
    将所述点数量小于阈值的聚类类别对应的所有点进行去除处理；
    选取去除处理后的聚类结果中各点对应的邻域点；及
    根据各点对应的邻域点计算相应点对应的法向量。
  8. 一种点云法向量计算装置，包括：
    获取模块，用于获取点云数据；
    分割模块，用于调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
    聚类模块，用于根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
    选取模块，用于根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
    计算模块，用于根据各点对应的邻域点计算相应点对应的法向量。
  9. 根据权利要求8所述的装置，其特征在于，所述聚类模块还用于根据所述语义分割结果对所述点云数据进行连通域检测，得到多个连通域；及根据所述连通域确定所述点云数据对应的多个聚类类别以及所述点云数据中各点对应的聚类类别，将所述聚类类别以及所述点云数据中各点对应的聚类类别作为聚类结果。
  10. 根据权利要求8所述的装置，其特征在于，所述选取模块，还用于获取所述点云数据中的当前点；根据所述聚类结果中确定与所述当前点属于相同聚类类别的同类点；在所述同类点中选取所述当前点对应的邻域点；及遍历所述点云数据中的所有点，得到所述点云数据中各点对应的邻域点。
  11. 一种计算机设备，包括存储器及一个或多个处理器，所述存储器中存储有计算机可读指令，所述计算机可读指令被所述一个或多个处理器执行时，使得所述一个或多个处理器执行以下步骤：
    获取点云数据；
    调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
    根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
    根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
    根据各点对应的邻域点计算相应点对应的法向量。
  12. 根据权利要求11所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：所述根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果包括：根据所述语义分割结果对所述点云数据进行连通域检测，得到多个连通域；及根据所述连通域确定所述点云数据对应的多个聚类类别以及所述点云数据中各点对应的聚类类别，将所述聚类类别以及所述点云数据中各点对应的聚类类别作为聚类结果。
  13. 根据权利要求11所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：获取所述点云数据中的当前点；根据所述聚类结果中确定与所述当前点属于相同聚类类别的同类点；在所述同类点中选取所述当前点对应的邻域点； 及遍历所述点云数据中的所有点，得到所述点云数据中各点对应的邻域点。
  14. 根据权利要求13所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：根据预设距离参数在所述同类点中选取预设数量的点，得到所述当前点对应的邻域点。
  15. 根据权利要求13所述的计算机设备，其特征在于，所述处理器执行所述计算机可读指令时还执行以下步骤：在所述同类点中选取与所述当前点在预设球体半径内的点，得到所述当前点对应的邻域点。
  16. 一个或多个存储有计算机可读指令的非易失性计算机可读存储介质，所述计算机可读指令被一个或多个处理器执行时，使得所述一个或多个处理器执行以下步骤：
    获取点云数据；
    调用预先训练的语义分割模型，将所述点云数据输入至所述语义分割模型，通过所述语义分割模型对所述点云数据进行语义分割，得到语义分割结果；
    根据所述语义分割结果对所述点云数据进行聚类，得到聚类结果；
    根据所述聚类结果选取所述点云数据中各点对应的邻域点；及
    根据各点对应的邻域点计算相应点对应的法向量。
  17. 根据权利要求16所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时还执行以下步骤：根据所述语义分割结果对所述点云数据进行连通域检测，得到多个连通域；及根据所述连通域确定所述点云数据对应的多个聚类类别以及所述点云数据中各点对应的聚类类别，将所述聚类类别以及所述点云数据中各点对应的聚类类别作为聚类结果。
  18. 根据权利要求16所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时还执行以下步骤：获取所述点云数据中的当前点；根据所述聚类结果中确定与所述当前点属于相同聚类类别的同类点；在所述同类点中选取所述当前点对应的邻域点；及遍历所述点云数据中的所有点，得到所述点云数据中各点对应的邻域点。
  19. 根据权利要求18所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时还执行以下步骤：根据预设距离参数在所述同类点中选取预设数量的点，得到所述当前点对应的邻域点。
  20. 根据权利要求18所述的存储介质，其特征在于，所述计算机可读指令被所述处理器执行时还执行以下步骤：在所述同类点中选取与所述当前点在预设球体半径内的点，得到所述当前点对应的邻域点。
PCT/CN2020/128259 2020-11-12 2020-11-12 点云法向量计算方法、装置、计算机设备和存储介质 WO2022099528A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2020/128259 WO2022099528A1 (zh) 2020-11-12 2020-11-12 点云法向量计算方法、装置、计算机设备和存储介质
CN202080092975.5A CN114930402A (zh) 2020-11-12 2020-11-12 点云法向量计算方法、装置、计算机设备和存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/128259 WO2022099528A1 (zh) 2020-11-12 2020-11-12 点云法向量计算方法、装置、计算机设备和存储介质

Publications (1)

Publication Number Publication Date
WO2022099528A1 true WO2022099528A1 (zh) 2022-05-19

Family

ID=81601943

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/128259 WO2022099528A1 (zh) 2020-11-12 2020-11-12 点云法向量计算方法、装置、计算机设备和存储介质

Country Status (2)

Country Link
CN (1) CN114930402A (zh)
WO (1) WO2022099528A1 (zh)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116246069A (zh) * 2023-02-07 2023-06-09 北京四维远见信息技术有限公司 自适应地形点云滤波的方法、装置、智能终端及存储介质
CN117252863A (zh) * 2023-11-13 2023-12-19 山东省地质测绘院 一种地理信息异常数据快速检测分析方法
CN117495891A (zh) * 2023-12-29 2024-02-02 法奥意威(苏州)机器人系统有限公司 点云边缘检测方法、装置和电子设备
CN117808703A (zh) * 2024-02-29 2024-04-02 南京航空航天大学 一种多尺度大型部件装配间隙点云滤波方法
CN117808703B (zh) * 2024-02-29 2024-05-10 南京航空航天大学 一种多尺度大型部件装配间隙点云滤波方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110274343A1 (en) * 2010-05-07 2011-11-10 Honeywell International, Inc. System and method for extraction of features from a 3-d point cloud
CN103985149A (zh) * 2014-05-20 2014-08-13 南京理工大学 三维彩色点云点特征描述方法
CN104123746A (zh) * 2014-07-10 2014-10-29 上海大学 一种三维扫描点云中实时法向量的计算方法
CN110992341A (zh) * 2019-12-04 2020-04-10 沈阳建筑大学 一种基于分割的机载LiDAR点云建筑物提取方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110274343A1 (en) * 2010-05-07 2011-11-10 Honeywell International, Inc. System and method for extraction of features from a 3-d point cloud
CN103985149A (zh) * 2014-05-20 2014-08-13 南京理工大学 三维彩色点云点特征描述方法
CN104123746A (zh) * 2014-07-10 2014-10-29 上海大学 一种三维扫描点云中实时法向量的计算方法
CN110992341A (zh) * 2019-12-04 2020-04-10 沈阳建筑大学 一种基于分割的机载LiDAR点云建筑物提取方法

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116246069A (zh) * 2023-02-07 2023-06-09 北京四维远见信息技术有限公司 自适应地形点云滤波的方法、装置、智能终端及存储介质
CN116246069B (zh) * 2023-02-07 2024-01-16 北京四维远见信息技术有限公司 自适应地形点云滤波的方法、装置、智能终端及存储介质
CN117252863A (zh) * 2023-11-13 2023-12-19 山东省地质测绘院 一种地理信息异常数据快速检测分析方法
CN117252863B (zh) * 2023-11-13 2024-02-09 山东省地质测绘院 一种地理信息异常数据快速检测分析方法
CN117495891A (zh) * 2023-12-29 2024-02-02 法奥意威(苏州)机器人系统有限公司 点云边缘检测方法、装置和电子设备
CN117495891B (zh) * 2023-12-29 2024-04-30 法奥意威(苏州)机器人系统有限公司 点云边缘检测方法、装置和电子设备
CN117808703A (zh) * 2024-02-29 2024-04-02 南京航空航天大学 一种多尺度大型部件装配间隙点云滤波方法
CN117808703B (zh) * 2024-02-29 2024-05-10 南京航空航天大学 一种多尺度大型部件装配间隙点云滤波方法

Also Published As

Publication number Publication date
CN114930402A (zh) 2022-08-19

Similar Documents

Publication Publication Date Title
WO2022099528A1 (zh) 点云法向量计算方法、装置、计算机设备和存储介质
US10699134B2 (en) Method, apparatus, storage medium and device for modeling lane line identification, and method, apparatus, storage medium and device for identifying lane line
JP6681729B2 (ja) オブジェクトの３ｄ姿勢およびオブジェクトのランドマーク点の３ｄロケーションを求める方法、およびオブジェクトの３ｄ姿勢およびオブジェクトのランドマークの３ｄロケーションを求めるシステム
CN106407947B (zh) 用于无人驾驶车辆的目标物体识别方法和装置
CN109977997B (zh) 基于卷积神经网络快速鲁棒的图像目标检测与分割方法
CN106886980B (zh) 一种基于三维激光雷达目标识别的点云密度增强的方法
CN110458772B (zh) 一种基于图像处理的点云滤波方法、装置和存储介质
WO2022099530A1 (zh) 点云数据运动分割方法、装置、计算机设备和存储介质
Cheng et al. Building boundary extraction from high resolution imagery and lidar data
CN109816051B (zh) 一种危化品货物特征点匹配方法及系统
CN113537180B (zh) 树障的识别方法、装置、计算机设备和存储介质
WO2022133770A1 (zh) 点云法向量的生成方法、装置、计算机设备和存储介质
CN104217433A (zh) 一种分析图像的方法及装置
JP2019536164A (ja) 画像処理装置、画像処理方法および画像処理プログラム
CN111951283A (zh) 一种基于深度学习的医学图像识别方法及系统
CN115457492A (zh) 目标检测方法、装置、计算机设备及存储介质
Zhao et al. Region-based saliency estimation for 3D shape analysis and understanding
CN113902782A (zh) 一种挖掘机周边障碍物三维点云的快速配准方法及系统
CN112330569A (zh) 模型训练方法、文本去噪方法、装置、设备及存储介质
CN110007764B (zh) 一种手势骨架识别方法、装置、系统及存储介质
CN114972339A (zh) 用于推土机结构件生产异常检测的数据增强系统
US11816857B2 (en) Methods and apparatus for generating point cloud histograms
CN112862730B (zh) 点云特征增强方法、装置、计算机设备和存储介质
CN116931583B (zh) 移动对象的确定方法和避让方法、装置、设备及存储介质
WO2022226831A1 (zh) 未定义类别障碍物的检测方法、装置和计算机设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20961085

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20961085

Country of ref document: EP

Kind code of ref document: A1