Point cloud data screening method and device and storage medium
By dividing point cloud data into three-dimensional block sets and filtering out target block units that meet the conditions, the problem of inefficient point cloud data screening in the existing technology is solved, and more efficient point cloud data annotation is achieved.
Patent Information
- Application Number
- CN202411033642.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-08
AI Technical Summary
The prior art requires full calculation of all point cloud data during the point cloud data screening process, resulting in inefficient labeling, especially in large orders of magnitude point cloud scenarios, resulting in delays and lags in the annotation tool.
By dividing point cloud data into a three-dimensional block set and filtering out target block units that meet the intersection conditions based on the labeling graph, the target point cloud data corresponding to the labeling graph is thus filtering out the calculation amount of the point cloud data screening process.
It improves the labeling efficiency of point cloud data, reduces the consumption of computing resources, reduces the time-consuming of point cloud screening process, and avoids the delay and lag of the labeling tool.
Smart Images

Figure CN118918262A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of data annotation, and in particular to a point cloud data screening method, device and storage medium. Background Art
[0002] In the field of computer vision and machine learning, with the development of sensor technologies such as lidar and depth cameras, high-precision three-dimensional point cloud data containing rich spatial information can be collected and generated. However, these initially acquired three-dimensional point cloud data usually lack structural and semantic information and need to be further semantically segmented and labeled.
[0003] In the field of data labeling, point cloud semantic segmentation and labeling aims to accurately segment point cloud data in different areas into subsets with different semantic labels, such as roads, vehicles, pedestrians, etc., based on the spatial geometric structure and shape information of point cloud data, and use computer vision algorithms to analyze and train models for these subsets and their labels, so that the trained recognition model can recognize and distinguish various elements in the current environment based on point cloud data, thereby building a detailed and complex environmental model.
[0004] The key to semantic segmentation and annotation of point clouds is to use annotation tools to select point clouds in a specific area in a complex 3D scene and annotate the semantic features of the selected point clouds. At present, the conventional point cloud screening method is: first project the 3D point cloud data onto a specified plane, draw a graph through the plane, and then traverse the 3D point cloud data to filter out the points that fall into the graph.
[0005] However, the above method needs to traverse and calculate each point cloud in the scene during the point cloud data screening process to determine whether each point cloud is selected. Therefore, especially for some large-scale point cloud scenes, the existing method for point cloud screening requires a lot of computing performance and time, resulting in delays and freezes in the annotation tool, and low annotation efficiency. Summary of the invention
[0006] The purpose of the embodiments of the present application is to provide a point cloud data screening method, device and storage medium to solve the problem of low labeling efficiency caused by the need to fully calculate all point cloud data during the point cloud screening process.
[0007] In a first aspect, an embodiment of the present application provides a point cloud data screening method, comprising:
[0008] Determine a set of three-dimensional blocks corresponding to the point cloud data to be annotated; wherein the set of three-dimensional blocks is composed of a plurality of three-dimensional block units;
[0009] Obtaining the annotation graphics drawn by the user based on the preset annotation plane;
[0010] Determining a three-dimensional collision recognition area corresponding to the marked graphic;
[0011] Filtering out a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set;
[0012] The target point cloud data corresponding to the annotated graphic is screened out based on the candidate point cloud data within the target block unit set.
[0013] In an embodiment of the present application, the spatial area occupied by the point cloud data is divided into a set of multiple three-dimensional blocks, and then the target blocks that meet the intersection conditions are screened out based on the labeled graphics, and then the target point cloud corresponding to the labeled graphics is screened out from the point cloud data surrounded by the target blocks, thereby greatly reducing the amount of calculation in the point cloud data screening process and improving the labeling efficiency of the point cloud data.
[0014] In some possible embodiments, determining a set of three-dimensional blocks corresponding to the point cloud data to be annotated includes:
[0015] Determine an original three-dimensional space region corresponding to the point cloud data to be annotated;
[0016] Based on a preset point cloud quantity threshold, the original three-dimensional space area is divided into an octree to obtain a set of three-dimensional blocks corresponding to the point cloud data to be annotated;
[0017] The three-dimensional block set is composed of a plurality of three-dimensional block units, and the number of point clouds contained in each of the three-dimensional block units is greater than the point cloud number threshold.
[0018] In the embodiment of the present application, the three-dimensional block set corresponding to the point cloud data to be labeled is determined by octree segmentation, which further improves the efficiency of obtaining the three-dimensional block set.
[0019] In some possible embodiments, determining the three-dimensional collision recognition area corresponding to the marked graphic includes:
[0020] A circumscribed circle corresponding to the marked figure is determined, and a three-dimensional collision recognition area corresponding to the marked figure is determined based on the circumscribed circle.
[0021] In the embodiment of the present application, the three-dimensional collision recognition area is determined by using the circumscribed circle of the annotated figure, thereby further improving the efficiency of the point cloud data screening process.
[0022] In some possible embodiments, determining the three-dimensional collision recognition area corresponding to the marked graphic based on the circumscribed circle includes:
[0023] A ray perpendicular to the annotation plane is constructed based on the center of the circumscribed circle, and an area within a radius of the circumscribed circle from the ray is determined as a three-dimensional collision recognition area.
[0024] In an embodiment of the present application, by constructing a ray perpendicular to the annotation plane based on the center of the circumscribed circle, and determining the area within the radius limit of the circumscribed circle as the three-dimensional collision identification area, the efficiency of the point cloud data screening process is further improved.
[0025] In some possible embodiments, the step of selecting a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set includes:
[0026] Determining an intersection distance threshold based on the side length of a single three-dimensional block unit and the collision radius of the three-dimensional collision identification area;
[0027] The vertical distance between the center point of each three-dimensional block unit and the ray is obtained, and the three-dimensional block units corresponding to the vertical distances less than the intersection distance threshold are screened out to form the target block unit set.
[0028] In the embodiment of the present application, by using the vertical distance between the center point of the cube and the ray to determine whether each cube intersects with the three-dimensional collision identification area, the efficiency of the point cloud data screening process is further improved.
[0029] In some possible embodiments, the step of filtering out target point cloud data corresponding to the annotated graphic based on candidate point cloud data within the target block unit set includes:
[0030] Determine first candidate point cloud data within the target block unit set, and select second candidate point cloud data that falls within the three-dimensional collision recognition area from the first candidate point cloud data;
[0031] Target point cloud data corresponding to the annotated graphic is screened out from the second candidate point cloud data.
[0032] In an embodiment of the present application, the efficiency of the point cloud data screening process is further improved by first screening out second candidate point cloud data that falls within the three-dimensional collision recognition area from the first candidate point cloud data within the target block unit range, and then screening the target point cloud data corresponding to the annotated graphic from the second candidate point cloud data.
[0033] In some possible embodiments, the step of selecting target point cloud data corresponding to the annotated graphic from the second candidate point cloud data includes:
[0034] Obtaining the two-dimensional coordinates of each candidate point in the second candidate point cloud data projected onto the annotation plane;
[0035] The target point cloud data is obtained by filtering out candidate points whose two-dimensional coordinates fall within the annotated figure from the second candidate point cloud data; or, the target point cloud data is obtained by filtering candidate points whose two-dimensional coordinates do not fall within the annotated figure in the second candidate point cloud data.
[0036] In an embodiment of the present application, the efficiency of the point cloud data screening process is further improved by obtaining the two-dimensional coordinate information of each point cloud of the second candidate point cloud data projected onto the annotation plane, and then comparing the two-dimensional coordinates with the annotation graphics to screen out the target point cloud data corresponding to the annotation graphics.
[0037] In a second aspect, an embodiment of the present application provides a point cloud data screening device, comprising:
[0038] A block determination module is used to determine a set of three-dimensional blocks corresponding to the point cloud data to be annotated; wherein the set of three-dimensional blocks is composed of a plurality of three-dimensional block units;
[0039] A graphic annotation module is used to obtain an annotation graphic drawn by a user based on a preset annotation plane;
[0040] An area determination module, used to determine a three-dimensional collision recognition area corresponding to the marked graphic;
[0041] A block screening module, used to screen out a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set;
[0042] The point cloud screening module is used to screen out target point cloud data corresponding to the annotated graphic based on the candidate point cloud data within the target block unit set.
[0043] In a third aspect, an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the method described in any embodiment of the first aspect when executing the program.
[0044] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any embodiment of the first aspect can be implemented.
[0045] In a fifth aspect, an embodiment of the present application provides a computer program product, wherein the computer program product includes a computer program, wherein the computer program, when executed by a processor, can implement the method described in any embodiment of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0047] Figure 1 A schematic diagram of a process of a point cloud data screening method provided in an embodiment of the present application;
[0048] Figure 2 A schematic diagram of the structure of a point cloud data screening device provided in an embodiment of the present application;
[0049] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0050] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.
[0051] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.
[0052] It should be noted that the current 3D point cloud annotation screening method mainly includes the following steps: 1. Specify a plane in the 3D scene; 2. Project the points in the scene onto the plane; 3. Use graphic tools to draw annotation graphics on the plane; 4. Traverse and filter out the points inside the drawn graphics as the selected points. However, the existing method requires traversal calculation and screening judgment of all points in the point cloud data. Especially for large-scale point cloud scenes, the amount of calculation is huge, which consumes a lot of system performance and computing time, resulting in freezes and delays in the annotation process, seriously affecting the overall annotation efficiency.
[0053] In response to the problems existing in the above-mentioned prior art, an embodiment of the present application provides a point cloud data screening method, which blocks the point cloud data, and then filters the point cloud from the target block after filtering out the target block according to the labeled graphics, thereby omitting the process of calculating and screening the point cloud surrounded by the remaining blocks, thereby greatly improving the labeling efficiency.
[0054] like Figure 1 As shown, the embodiment of the present application provides a point cloud data screening method, which may include the steps of:
[0055] S1. Determine a 3D block set corresponding to the point cloud data to be annotated; wherein the 3D block set is composed of a plurality of 3D block units.
[0056] It should be noted that by using a laser radar or other equipment to scan a scene of interest, point cloud data of the scene of interest can be generated. After preprocessing these initially generated point cloud data (such as filtering, boundary determination and cropping), point cloud data to be annotated is obtained. The next step is to select 3D objects and annotate semantic features for the point cloud data to be annotated, such as annotating vehicles, pedestrians, obstacles, etc. in the driving scene.
[0057] Specifically, the point cloud data to be annotated is a collection of a series of points with three-dimensional spatial coordinate information. Based on the spatial coordinate information of these points, a three-dimensional block set can be established. This three-dimensional block set generally depicts the three-dimensional spatial area occupied by the point cloud data. The three-dimensional block set is composed of a number of three-dimensional block units. Each three-dimensional block unit is not limited to a cube or a cuboid, but for the convenience of calculation, the three-dimensional block unit is usually a cube.
[0058] Exemplarily, a minimum circumscribed cube that encloses all the point cloud data to be annotated can be first generated, and then the minimum circumscribed cube can be divided into a number of small cubes (such as cubes) according to a preset unit side length, and then the number of point clouds enclosed by each small cube is counted, and the small cubes with less than the preset number of point clouds are filtered out, and the remaining small cubes are the three-dimensional cube sets corresponding to the point cloud data to be annotated. It should be noted that if the number of point clouds contained in a small cube is too small, it is regarded as noise or an unimportant spatial area; if the generated point cloud data to be annotated is relatively accurate, the preset number can also be set to 0, that is, all small cubes that do not contain point cloud data are filtered out.
[0059] S2. Obtain the annotation graphic drawn by the user based on a preset annotation plane.
[0060] Specifically, a marking plane can be selected according to the actual marking requirements. Usually, a mathematical plane perpendicular to the current 3D camera perspective is selected as the marking plane. It can be understood that the current 3D camera perspective refers to a coordinate system constructed with the camera as the origin and the front of the perspective as the positive direction. All the currently collected point cloud data are assigned spatial coordinates in this coordinate system.
[0061] It can be understood that the user can see the three-dimensional scene displayed by the point cloud data to be annotated based on the current 3D camera perspective. According to the distribution of the point cloud data in the scene, the user can draw the desired annotation graphics based on the annotation plane, such as circles, squares, or irregular graphics such as vehicle shapes.
[0062] S3. Determine a three-dimensional collision recognition area corresponding to the marked graphic.
[0063] Specifically, based on the outline of the annotated figure extending in a direction perpendicular to the annotated plane (the extension distance can be set according to the longest distance of the point cloud data), a cylindrical three-dimensional area with a cross-section of the annotated figure can be formed, that is, a three-dimensional collision recognition area corresponding to the annotated figure.
[0064] It should be noted that when determining the three-dimensional collision recognition area, it is not limited to extending the outline of the annotated figure in a direction perpendicular to the annotated plane, but can also be extending the outline of a circumscribed circle, circumscribed rectangle or other figure of the annotated figure in a direction perpendicular to the annotated plane to form a three-dimensional collision recognition area corresponding to the annotated figure. The specific setting can be based on actual annotation requirements.
[0065] S4. Filter out a target block unit set intersecting with the three-dimensional collision recognition area from the three-dimensional block set.
[0066] Specifically, based on the determined three-dimensional collision recognition area, the target block unit set can be screened out according to the intersection of each three-dimensional block unit in the three-dimensional block set with the three-dimensional collision recognition area (ie, the occurrence of "collision").
[0067] It can be understood that the three-dimensional block unit can be represented as a spatial coordinate range, and the three-dimensional collision identification area can also be represented as a spatial coordinate range. According to the intersection operation algorithm between the spatial coordinate ranges, it can be determined whether the two three-dimensional areas intersect. Exemplarily, judging whether each three-dimensional block unit intersects with the three-dimensional collision identification area can be understood as judging whether there is at least one point of the coordinates of a single three-dimensional block unit that falls within the spatial coordinate range of the three-dimensional collision identification area. By judging and screening the three-dimensional block units that intersect with the three-dimensional collision identification area one by one, a target block unit set is formed.
[0068] S5. Filter out target point cloud data corresponding to the annotated graphic based on the candidate point cloud data within the target block unit set.
[0069] Specifically, first determine which points in the point cloud data to be annotated belong to the candidate point cloud data within the range of the target block unit set, and then filter out the target point cloud data corresponding to the annotated figure from the points (candidate point cloud data) contained in the target block unit set.
[0070] Exemplarily, the spatial coordinate range corresponding to each target block unit can be determined respectively, and then the points surrounded by the target block unit can be directly filtered out from the point cloud data to be annotated based on the spatial coordinate range, and so on to obtain candidate point cloud data within the target block unit set range.
[0071] Then, the target point cloud data corresponding to the annotated graphics is filtered out from the candidate point cloud data. For example, based on the spatial coordinate range of the stereoscopic collision recognition area determined above, the target point cloud data falling within the spatial coordinate range can be filtered out from the candidate point cloud data. After the target point cloud data is selected, the target point cloud data can be added as a target object, and the corresponding semantic feature annotation information can be added.
[0072] Based on this, the existing method of calculating and screening all point cloud data is changed to a method of first screening the target blocks and then calculating and screening the point cloud data, thereby greatly reducing the amount of calculation in the point cloud screening process, and being able to quickly select the target point cloud data to be labeled, effectively improving the labeling efficiency of the point cloud data.
[0073] In some possible embodiments, in step S1, determining a set of three-dimensional blocks corresponding to the point cloud data to be annotated may include:
[0074] S101, determining an original three-dimensional space region corresponding to the point cloud data to be annotated;
[0075] S102, performing octree segmentation on the original three-dimensional space region based on a preset point cloud quantity threshold to obtain a set of three-dimensional blocks corresponding to the point cloud data to be annotated;
[0076] The three-dimensional block set is composed of a plurality of three-dimensional block units, and the number of point clouds contained in each three-dimensional block unit is greater than a point cloud number threshold.
[0077] It should be noted that, in addition to the above method of first dividing the three-dimensional block units and then filtering according to the point cloud data contained in each unit, the three-dimensional block set corresponding to the point cloud data to be labeled can also be determined by octree segmentation.
[0078] Specifically, first determine an original three-dimensional space area corresponding to the entire point cloud data to be annotated, and then divide the original three-dimensional space area into 8 identical small three-dimensional space areas. If the number of point clouds contained in these small three-dimensional space areas exceeds the preset number, they are eliminated if not, and if they exceed, the small three-dimensional space area is used as the basis for further division into 8 identical smaller three-dimensional space areas, and so on, until it is divided into the smallest three-dimensional space area (that is, the final three-dimensional block unit). Finally, a three-dimensional block set composed of several three-dimensional block units is obtained. It can be understood that since the three-dimensional blocks containing insufficient point cloud data have been eliminated and filtered in the step-by-step segmentation process, the number of point clouds contained in each of the remaining three-dimensional block units is greater than the preset point cloud number threshold.
[0079] In some possible embodiments, step S3, determining a three-dimensional collision recognition area corresponding to the marked graphic, may include:
[0080] S301 , determining a circumscribed circle corresponding to the marked figure, and determining a three-dimensional collision recognition area corresponding to the marked figure based on the circumscribed circle.
[0081] It should be noted that when the marked figure is an irregular figure, in order to simplify calculation, the three-dimensional collision recognition area is usually not determined directly by the outline of the marked figure, but by using the circumscribed circle, circumscribed rectangle, etc. of the marked figure as the outline to determine the three-dimensional collision recognition area.
[0082] Exemplarily, an embodiment of the present application first determines the circumscribed circle of the annotated figure, and then extends the outline of the circumscribed circle in a direction perpendicular to the annotated plane (the outline of the figure can be approximately represented as a set of a finite number of points, and then based on these points, they are respectively extended in a direction perpendicular to the annotated plane, that is, a corresponding finite number of rays are formed) to form a cylindrical spatial region (for example, the region enclosed by the aforementioned finite number of rays can be approximately regarded as a cylindrical spatial region), and this cylindrical spatial region is the three-dimensional collision recognition region corresponding to the annotated figure.
[0083] In some possible embodiments, in step S301, determining the three-dimensional collision recognition area corresponding to the marked graphic based on the circumscribed circle may include:
[0084] S3011. Construct a ray perpendicular to the annotation plane based on the center of the circumscribed circle, and determine an area within the radius of the circumscribed circle from the ray as a three-dimensional collision recognition area.
[0085] For example, first, a ray perpendicular to the annotation plane is constructed based on the center of the circumscribed circle (the length can be set according to the longest distance of the point cloud data to be annotated), and then the ray is used as the central axis. At each point on the central axis (the line segment can be approximately represented as a set of a finite number of points, and the specific density of the points can be set according to requirements), a circle is drawn around the central axis based on the radius of the circumscribed circle, and the corresponding spatial area surrounded by multiple circles is obtained, which is the three-dimensional collision recognition area corresponding to the annotated figure. In addition, the three-dimensional collision recognition area corresponding to the annotated figure can also be directly constructed according to the cylinder representation algorithm in some 3D annotation tools.
[0086] In some possible embodiments, step S4, screening out a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set, may include:
[0087] S401, determining an intersection distance threshold based on the side length of a single three-dimensional block unit and the collision radius of a three-dimensional collision identification area;
[0088] S402 , obtaining the vertical distance between the center point of each 3D block unit and the ray, and filtering out the 3D block units corresponding to the vertical distances smaller than the intersection distance threshold to form a target block unit set.
[0089] It should be noted that the collision radius of the three-dimensional collision recognition area can be the radius of the circumscribed circle of the annotated figure, and the intersection distance threshold can be the sum of the side length of a single three-dimensional block unit and the collision radius. It can be understood that when the center of one side of a certain three-dimensional block unit just touches the three-dimensional collision recognition area, the distance between the center point of the body of the three-dimensional block unit and the central axis of the three-dimensional collision recognition area (i.e., the above-mentioned ray) is exactly the intersection distance threshold (i.e., the sum of the side length of a single three-dimensional block unit and the collision radius); then, using the intersection distance threshold as the judgment condition, when the vertical distance between the center point of the three-dimensional block unit and the ray is less than the intersection distance threshold, it means that the three-dimensional block unit intersects with the three-dimensional collision recognition area, and is screened as a target block unit. By analogy, each three-dimensional block unit of the three-dimensional block set is judged one by one, and finally the target block unit set is screened.
[0090] In addition, the eight vertex coordinate values of each three-dimensional block unit may be directly obtained to determine whether these vertex coordinates fall within the range of the three-dimensional collision recognition area, so as to approximately determine whether the three-dimensional block unit intersects with the three-dimensional collision recognition area.
[0091] In some possible embodiments, step S5, filtering out target point cloud data corresponding to the annotated graphic based on candidate point cloud data within the target block unit set range, may include:
[0092] S501, determining first candidate point cloud data within the target block unit set range, and screening out second candidate point cloud data falling within the three-dimensional collision recognition area from the first candidate point cloud data;
[0093] S502: Filter out target point cloud data corresponding to the annotated graphic from the second candidate point cloud data.
[0094] It should be noted that since the cross-section of the three-dimensional collision recognition area is not necessarily formed with the annotated figure as the outline, for example, the annotated figure is a polygon and the three-dimensional collision recognition area is determined based on the circumscribed circle of the polygon, the target point cloud data screening can be performed in two steps: 1. First, the second candidate point cloud data falling within the three-dimensional collision recognition area is screened out from the first candidate point cloud data within the target block unit set; 2. The target point cloud data corresponding to the annotated figure is further screened out from the second candidate point cloud data.
[0095] It should be noted that step S5 can first determine whether the cross-section of the three-dimensional collision recognition area parallel to the annotation plane is the same figure as the annotation figure (for example, both are circles): if so, then directly filter out the points falling into the three-dimensional collision recognition area from the first candidate point cloud data within the target block unit set range, which is the final target point cloud data; if not, it is necessary to implement it step by step according to the above two steps S501 to S502.
[0096] In some possible embodiments, step S502, screening out target point cloud data corresponding to the annotated graphic from the second candidate point cloud data, may include:
[0097] S5021, obtaining the two-dimensional coordinates of each candidate point in the second candidate point cloud data projected onto the annotation plane;
[0098] S5022. Filter out candidate points whose two-dimensional coordinates fall within the annotated graphics from the second candidate point cloud data to obtain target point cloud data; or filter candidate points whose two-dimensional coordinates do not fall within the annotated graphics in the second candidate point cloud data to obtain target point cloud data.
[0099] Exemplarily, when the cross-section of the three-dimensional collision recognition area parallel to the annotation plane is a different figure from the annotation figure, for example, the former is a circle and the latter is a rectangle, step S501 has filtered out the second candidate point cloud data that falls within the three-dimensional collision recognition area from the first candidate point cloud data, that is, the filtered out second candidate point cloud data are points within the cylindrical range corresponding to the three-dimensional collision recognition area, and it is necessary to further filter out the points corresponding to the annotation figure (that is, points within the three-dimensional space range with a rectangular cross-section) from these second candidate point cloud data.
[0100] Specifically, the two-dimensional coordinates of each candidate point in the second candidate point cloud data projected onto the annotation plane can be calculated first. At this time, it is equivalent to converting the range judgment of the three-dimensional point into the range judgment of the two-dimensional point. According to whether the two-dimensional coordinates of each candidate point fall within the range of the annotation figure (the current instance is a rectangle), it is judged whether the candidate point is the target point cloud. Among them, step S5022 can be implemented in two ways: 1. The candidate points that fall within the range of the annotation figure are filtered as the target point cloud data; 2. The candidate points that do not fall within the range of the annotation figure are eliminated and filtered, and the remaining candidate points are the target point cloud data.
[0101] It should be noted that, after testing, in the process of annotating scenes with the same number of point clouds, the time consumed for point cloud screening using the method of the embodiment of the present application is much less than the time consumed for point cloud screening using the existing conventional scheme.
[0102] It can be understood that since the embodiment of the present application first performs block screening and then performs point cloud screening, there is no need to render and calculate the point cloud data that does not belong to the target block, which saves a lot of computing resources and time, and greatly improves the efficiency of 3D point cloud data annotation.
[0103] Please refer to Figure 2 , Figure 2 FIG. 1 is a block diagram showing the structure of a point cloud data screening device provided by some embodiments of the present application. It should be understood that the point cloud data screening device is similar to the above-mentioned Figure 1 Corresponding to the method embodiment, it is able to execute each step involved in the above method embodiment. The specific functions of the point cloud data screening device can be found in the description above. To avoid repetition, the detailed description is appropriately omitted here.
[0104] Figure 2 The point cloud data screening device includes at least one software function module that can be stored in a memory in the form of software or firmware or solidified in the point cloud data screening device, and the point cloud data screening device includes:
[0105] The block determination module 210 is used to determine a set of three-dimensional blocks corresponding to the point cloud data to be annotated; wherein the set of three-dimensional blocks is composed of a plurality of three-dimensional block units;
[0106] The graphic annotation module 220 is used to obtain the annotation graphic drawn by the user based on the preset annotation plane;
[0107] The area determination module 230 is used to determine the three-dimensional collision recognition area corresponding to the marked graphic;
[0108] The block screening module 240 is used to screen out a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set;
[0109] The point cloud screening module 250 is used to screen out target point cloud data corresponding to the annotated graphic based on the candidate point cloud data within the target block unit set.
[0110] It can be understood that the above-mentioned device item embodiment corresponds to the method item embodiment of the present invention. A point cloud data screening device provided by the embodiment of the present invention can implement the point cloud data screening method provided by any method item embodiment of the present invention.
[0111] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method, and will not be described in detail here.
[0112] like Figure 3As shown, some embodiments of the present application provide an electronic device 300, which includes: a memory 310, a processor 320, and a computer program stored in the memory 310 and executable on the processor 320, wherein the processor 320 reads the program from the memory 310 through a bus 330 and executes the program to implement a method of any embodiment included in the above-mentioned point cloud data screening method.
[0113] Processor 320 can process digital signals and can include various computing structures, such as complex instruction set computer structure, reduced instruction set computer structure, or a structure that implements a combination of multiple instruction sets. In some examples, processor 320 can be a microprocessor.
[0114] The memory 310 may be used to store instructions executed by the processor 320 or data related to the execution of instructions. These instructions and / or data may include codes for implementing some or all functions of one or more modules described in the embodiments of the present application. The processor 320 of the disclosed embodiment may be used to execute instructions in the memory 310 to implement the method shown above. The memory 310 includes a dynamic random access memory, a static random access memory, a flash memory, an optical memory, or other memory known to those skilled in the art.
[0115] Some embodiments of the present application further provide a computer-readable storage medium having a computer program stored thereon. The computer program is executed by a processor to execute the method described in the method embodiment.
[0116] Some embodiments of the present application further provide a computer program product, which, when executed on a computer, enables the computer to execute the method described in the method embodiment.
[0117] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0118] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.
[0119] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
[0120] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned 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 disk.
[0121] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.
[0122] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0123] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
Claims
1. A point cloud data screening method, It is characterized in that include: Determine a set of three-dimensional blocks corresponding to the point cloud data to be annotated; wherein the set of three-dimensional blocks is composed of a plurality of three-dimensional block units; Obtaining the annotation graphics drawn by the user based on the preset annotation plane; Determining a three-dimensional collision recognition area corresponding to the marked graphic; Filtering out a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set; The target point cloud data corresponding to the annotated graphic is screened out based on the candidate point cloud data within the target block unit set.
2. The point cloud data screening method according to claim 1, It is characterized in that The step of determining a set of three-dimensional blocks corresponding to the point cloud data to be annotated includes: Determine an original three-dimensional space region corresponding to the point cloud data to be annotated; Based on a preset point cloud quantity threshold, the original three-dimensional space area is divided into an octree to obtain a set of three-dimensional blocks corresponding to the point cloud data to be annotated; The three-dimensional block set is composed of a plurality of three-dimensional block units, and the number of point clouds contained in each of the three-dimensional block units is greater than the point cloud number threshold.
3. The point cloud data screening method according to claim 1, It is characterized in that The determining of the three-dimensional collision recognition area corresponding to the marked graphic includes: A circumscribed circle corresponding to the marked figure is determined, and a three-dimensional collision recognition area corresponding to the marked figure is determined based on the circumscribed circle.
4. The point cloud data screening method according to claim 3, It is characterized in that The determining of the three-dimensional collision recognition area corresponding to the marked graphic based on the circumscribed circle includes: A ray perpendicular to the annotation plane is constructed based on the center of the circumscribed circle, and an area within a radius of the circumscribed circle from the ray is determined as a three-dimensional collision recognition area.
5. The point cloud data screening method according to claim 4, It is characterized in that The step of selecting a target block unit set intersecting with the three-dimensional collision identification area from the three-dimensional block set includes: Determining an intersection distance threshold based on the side length of a single three-dimensional block unit and the collision radius of the three-dimensional collision identification area; The vertical distance between the center point of each three-dimensional block unit and the ray is obtained, and the three-dimensional block units corresponding to the vertical distances less than the intersection distance threshold are screened out to form the target block unit set.
6. The point cloud data screening method according to claim 1, It is characterized in that The step of filtering out target point cloud data corresponding to the annotated graphic based on the candidate point cloud data within the target block unit set includes: Determine first candidate point cloud data within the target block unit set, and select second candidate point cloud data that falls within the three-dimensional collision recognition area from the first candidate point cloud data; Target point cloud data corresponding to the annotated graphic is screened out from the second candidate point cloud data.
7. The point cloud data screening method according to claim 6, It is characterized in that The step of selecting target point cloud data corresponding to the annotated graphic from the second candidate point cloud data includes: Obtaining the two-dimensional coordinates of each candidate point in the second candidate point cloud data projected onto the annotation plane; The target point cloud data is obtained by filtering out candidate points whose two-dimensional coordinates fall within the annotated graphic from the second candidate point cloud data; or, the target point cloud data is obtained by filtering candidate points whose two-dimensional coordinates do not fall within the annotated graphic in the second candidate point cloud data.
8. An electronic device, It is characterized in that The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the point cloud data screening method according to any one of claims 1 to 7 can be implemented when the processor executes the program.
9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the point cloud data screening method according to any one of claims 1 to 7 is executed.
10. A computer program product, It is characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the point cloud data screening method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Point cloud local quality inspection marking method and system, program product, equipment and medium
CN120783344A
Point cloud local quality inspection marking method, system, program product, device and medium
CN120783344B
