CN116245943A - Continuous frame point cloud data labeling method and device based on web - Google Patents
Continuous frame point cloud data labeling method and device based on web Download PDFInfo
- Publication number
- CN116245943A CN116245943A CN202211634344.9A CN202211634344A CN116245943A CN 116245943 A CN116245943 A CN 116245943A CN 202211634344 A CN202211634344 A CN 202211634344A CN 116245943 A CN116245943 A CN 116245943A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- frame
- annotation
- data
- cloud data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a web-based continuous frame point cloud data labeling method and a web-based continuous frame point cloud data labeling device, wherein the method comprises the following steps: acquiring an annotation request, responding to the annotation request, downloading continuous frame point cloud data to be annotated and corresponding 2D image data, and storing the continuous frame point cloud data and the corresponding 2D image data in a local database; acquiring an initial annotation frame of an object in one frame of point cloud data in continuous frame point cloud data, and automatically calculating based on the initial annotation frame to obtain a final annotation frame; and mapping the final annotation frame to the 2D image data. The method and the device can realize multi-frame big data loading and automatic convergence of the marking frame, improve marking efficiency, reduce marking difficulty, enable marking personnel to finish continuous big data marking of a specific scene in continuous time, improve data marking precision and accelerate production of marking data.
Description
Technical Field
The present invention relates generally to the field of autopilot, and more particularly to web-based continuous frame point cloud data annotation methods and apparatus.
Background
Along with the healthy development of the automatic driving industry, the automatic driving test road is continuously put off in various places, and related systems are continuously perfected, so that a good foundation is laid for the development of intelligent automobiles in China. The core module in the automatic driving field is a perception layer, a decision layer and a control layer, and mainly solves the core problem of automatic driving: what is i? What is i going? How does i arrive? The obstacle detection is an important content in the perception layer, and is a precondition for realizing an automatic driving function. The method utilizes various sensors to acquire the environmental information of the automobile and the information of surrounding vehicles, pedestrians, traffic lights, road signs and the like, captures 2D visual data through a camera, captures 3D position data through a laser radar, a millimeter wave radar and the like, provides data support for the comprehensive decision of the automobile, and is a core measure for solving the problem of 'I' where.
The current artificial intelligence is relatively mature in terms of calculation power and algorithm, but if specific pain points in the industry are required to be solved through the algorithm and the applied landing, a large amount of automatic driving scene data are required to be collected, and algorithm training support is performed after data processing and manual labeling. In the field of automatic driving annotation, a common annotation mode is to fuse single-frame 2D and 3D data of a web end, namely, annotate images and point cloud data acquired by a 2D (camera) and a 3D sensor (radar) at the same time. However, the labeling scheme of the single-frame small point cloud is insufficient for algorithm training of all scenes of automatic driving, and a method for supporting web-end large point cloud labeling is needed to be established for constructing a training data set and improving the capabilities of algorithms such as target tracking, semantic segmentation and the like.
Therefore, in order to solve the above problems, a novel web-based continuous frame point cloud data labeling method and device are needed to solve the problems of the existing point cloud labeling scheme.
Disclosure of Invention
In the summary, a series of concepts in a simplified form are introduced, which will be further described in detail in the detailed description. The summary of the invention is not intended to define the key features and essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.
According to an aspect of the invention, there is provided a web-based continuous frame point cloud data labeling method, the method comprising: acquiring a labeling request, responding to the labeling request, downloading continuous frame point cloud data to be labeled and corresponding 2D image data, and storing the continuous frame point cloud data and the corresponding 2D image data in a local database; acquiring an initial annotation frame of an object in one frame of point cloud data in the continuous frame of point cloud data, and automatically calculating based on the initial annotation frame to obtain a final annotation frame; and mapping the final annotation box to the 2D image data.
In one embodiment, the method further comprises: and downloading the continuous frame point cloud data and the 2D image data to be annotated and acquiring the initial annotation frame are simultaneously executed by using different processing resources.
In one embodiment, the local database is a browser's local database.
In one embodiment, wherein the final annotation frame is automatically calculated based on the initial annotation frame, comprising: acquiring point cloud coordinates of a point cloud set in the initial annotation frame; and calculating the minimum circumscribed rectangle of the object based on the point cloud coordinates, and automatically modifying the initial annotation frame according to the minimum circumscribed rectangle to obtain the final annotation frame.
In one embodiment, wherein calculating the minimum bounding rectangle of the object based on the point cloud coordinates comprises: estimating a height of the object based on the point cloud coordinates, and calculating an updated set of point clouds within the initial annotation frame based on the estimated height of the object; computing an image representation of the object using the updated set of point clouds; and calculating the size of the minimum bounding rectangle based on the image representation of the object, thereby obtaining the minimum bounding rectangle.
In one embodiment, the method further comprises: the point cloud coordinates are transformed from a radar coordinate system to a point cloud coordinate system before estimating the height of the object based on the point cloud coordinates.
In one embodiment, wherein estimating the height of the object based on the point cloud coordinates comprises: calculating the height difference between the highest point and the lowest point in the point cloud set; and estimating a height of the object by removing ground level from the level difference, wherein calculating an updated set of point clouds within the initial annotation frame based on the estimated height of the object comprises: points outside the estimated height range of the object are removed from the set of point clouds to obtain an updated set of point clouds.
In one embodiment, wherein computing the image representation of the object using the updated set of point clouds comprises: transforming the updated set of point clouds to an image coordinate system and calculating an image representation of the object based on the transformed set of point clouds.
In one embodiment, the method further comprises: and after the continuous frame point cloud data is marked, carrying out multi-frame fitting playback on the marked result.
According to another aspect of the present invention, there is provided a web-based continuous frame point cloud data labeling apparatus, the apparatus comprising a memory and a processor, the memory having stored thereon a computer program to be run by the processor, which when run by the processor causes the processor to perform the web-based continuous frame point cloud data labeling method as described above.
According to yet another aspect of the present invention, there is provided a computer readable medium having stored thereon computer executable instructions which, when executed, perform the web-based continuous frame point cloud data annotation method as described above.
According to the web-based continuous frame point cloud data labeling method and device, multi-frame big data loading can be achieved, labeling frames can automatically converge, labeling efficiency is improved, labeling difficulty is reduced, labeling personnel can finish continuous big data labeling of a specific scene in continuous time, data labeling precision is improved, and production of labeled data is accelerated.
Drawings
The following drawings are included to provide an understanding of the invention and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the invention and their description to explain the principles of the invention.
In the accompanying drawings:
FIG. 1 shows an exemplary step flow diagram of a web-based continuous frame point cloud data annotation method according to one embodiment of the invention;
FIG. 2 shows a schematic diagram of an example image representation according to an embodiment of the invention; and
FIG. 3 shows a schematic block diagram of a web-based continuous frame point cloud data annotation device according to one embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present invention and not all embodiments of the present invention, and it should be understood that the present invention is not limited by the example embodiments described herein. Based on the embodiments of the invention described in the present application, all other embodiments that a person skilled in the art would have without inventive effort shall fall within the scope of the invention.
As described above, the existing 2D/3D fusion labeling is based on point cloud data and 2D image data acquired by a radar and a camera, and is divided into a group of data according to frame correspondence, after a labeling client pulls data, the point cloud data and the 2D image data are loaded respectively, labeling such as frame pulling and labeling is performed on the point cloud on a labeling page or the client, and the labeling frame is mapped to the 2D image through transformation of a radar coordinate system and a camera coordinate system so as to assist in accurate labeling.
Specifically, the data annotation process for 2D/3D fusion annotation can be generally divided into the following steps:
step one: data loading, namely pulling the content of the fusion annotation data and the annotation parameters;
step two: loading and rendering a point cloud labeling area;
step three: loading 2D image information and displaying image data acquired by each camera;
step four: the 3D point Yun Lakuang is marked, and the drawn marked frame is projected to a 2D image through transformation of a radar coordinate system and a camera coordinate system;
step five: and storing the labeling result, and exporting the labeling result to an automatic driving model engine for training.
The existing 2D/3D fusion labeling scheme mainly adopts the steps of extracting key frames after carrying out point cloud analysis and video frame extraction and time sequence alignment on original data, and carrying out single-frame labeling, wherein although automatic driving data fusion labeling is carried out relatively simply, single frames are not connected, so that the labeling of the same target object is not easy to judge, and the reconstructed point cloud and continuous frame point cloud loading and labeling capability are lacked aiming at scenes such as target tracking, semantic segmentation and the like, so that on one hand, the data precision is insufficient, and on the other hand, the data labeling difficulty and cost are increased.
In summary, the current point cloud labeling method for the web end has the following problems:
1. the single-frame labeling mode causes no connection between frames, the labeling process needs to continuously switch point clouds so as to judge whether the labeled objects are the same object, and the labeling efficiency is low.
2. And the single-frame point cloud is switched back and forth, so that the data loading efficiency is low.
3. Aiming at the situations of complex urban road scenes, different vehicle orientations and high concentration, the cost of manually marking data is high at present.
4. Because the point cloud data in unit time are associated with time sequence, the single-frame labeling mode for extracting key frames at present cannot guarantee the accuracy and quality of the data.
Therefore, in order to solve the above-mentioned problems of the existing point cloud data labeling method, the present invention provides a web-based continuous frame point cloud data labeling method, which includes: acquiring a labeling request, responding to the labeling request, downloading continuous frame point cloud data to be labeled and corresponding 2D image data, and storing the continuous frame point cloud data and the corresponding 2D image data in a local database; acquiring an initial annotation frame of an object in one frame of point cloud data in the continuous frame of point cloud data, and automatically calculating based on the initial annotation frame to obtain a final annotation frame; and mapping the final annotation box to the 2D image data.
The web-based continuous frame point cloud data labeling method can realize multi-frame big data loading, automatically converge labeling frames, improve labeling efficiency, reduce labeling difficulty, enable labeling personnel to finish continuous big data labeling of specific scenes in continuous time, improve data labeling precision and accelerate labeling data production.
The following describes the web-based continuous frame point cloud data labeling method and the web-based continuous frame point cloud data labeling device according to the invention in detail by combining with a specific embodiment.
Referring initially to FIG. 1, FIG. 1 shows an exemplary flowchart of steps of a web-based continuous frame point cloud data annotation method 100 according to one embodiment of the invention.
As shown in fig. 1, the web-based continuous frame point cloud data annotation method 100 may include the following exemplary steps:
in step S110, a labeling request is acquired, and continuous frame point cloud data to be labeled and corresponding 2D image data are downloaded and stored in a local database in response to the labeling request;
in step S120, an initial labeling frame of an object in one frame of point cloud data in continuous frame point cloud data is obtained, and a final labeling frame is automatically calculated based on the initial labeling frame; and
in step S130, the final annotation frame is mapped to the 2D image data.
In one embodiment, the annotation request may be obtained by the user clicking on a certain annotation task.
In an actual labeling scene, the web side is very resource-consuming to load point cloud data and 2D image data with large data volume, and if all the resources are completely loaded, waiting time is too long. Thus, in one embodiment, the method 100 may further comprise: downloading the continuous frame point cloud data and the 2D image data to be annotated is performed simultaneously with acquiring the initial annotation frame using different processing resources. In one embodiment, web workbench provided by HTML 5 can be used to create a multithreading environment for JavaScript, allowing a main thread to create a workbench sub-thread, and distributing time-consuming operations of loading and analyzing residual point clouds in the main thread to the workbench sub-thread, so that after loading a first frame point cloud, the main thread can start processing labeling and rendering tasks, and meanwhile, the workbench sub-thread loads and analyzes the residual point clouds, and the two do not interfere with each other. After the processing of the workbench sub-thread is finished, the processing result is returned to the main thread through a message mechanism. The method and the device fully utilize the advantages of multithreading, load big data has no influence on the labeling process, can normally process operations such as rendering, clicking by a user and the like, and ensure that the main thread is not blocked and labeling and rendering pages are not blocked.
Because successive frames require batches of point cloud data, whereas a single point cloud contains hundreds of thousands of levels of point data, batch storage is very memory-intensive, and thus, in one embodiment, the local database may be a browser's local database, such as an IndexdDB, that is suitable for large data storage. Compared to cookies and local Storage (local Storage), the IndexdDB has the following properties:
a) Asynchronous: the indeeddb operation does not lock the browser and the user can still do other operations, in contrast to local storage, which is synchronized. The use of the IndexedDB does not affect the performance of the browser when storing data.
b) The storage space is large: the Storage space of the indireddb is much larger than the Local Storage, generally not less than 250MB, even without an upper limit.
c) Support binary storage: the index db may store not only character strings but also binary data (e.g., arrayBuffer objects and Blob objects).
According to the method, the pulled point cloud and the image data are stored in the IndexdDB, the data cached in the IndexdDB can be directly read when the page is refreshed or reloaded for the second time, the loading time caused by a secondary network request is reduced, and the labeling efficiency after the second loading is improved.
In one embodiment, the automatic calculation of the final annotation frame based on the initial annotation frame in step S120 may include the following steps:
acquiring point cloud coordinates of a point cloud set in an initial annotation frame; and
and calculating the minimum circumscribed rectangle of the object based on the point cloud coordinates, and automatically modifying the initial annotation frame according to the minimum circumscribed rectangle to obtain a final annotation frame.
Since the point cloud data is acquired by a radar (e.g., a laser radar, a millimeter wave radar, etc.), etc., the point cloud coordinates of the point cloud set within the initial annotation frame are coordinates of the radar coordinate system.
In one embodiment, calculating the minimum bounding rectangle of the object based on the point cloud coordinates may include the steps of:
estimating a height of the object based on the point cloud coordinates, and calculating an updated set of point clouds within the initial annotation frame based on the estimated height of the object;
computing an image representation of the object using the updated set of point clouds; and
the size of the minimum bounding rectangle is calculated based on the image representation of the object, resulting in a minimum bounding rectangle.
For ease of computation, in one embodiment, the point cloud coordinates are transformed from a radar coordinate system to a point cloud coordinate system before estimating the height of the object based on the point cloud coordinates. Specifically, a center point of the initial annotation frame may be selected as a coordinate origin, and coordinates of each point of the point cloud with respect to the center point may be calculated, thereby transforming the point cloud coordinates from the radar coordinate system to the point cloud coordinate system.
In one embodiment, wherein estimating the height of the object based on the point cloud coordinates may comprise: calculating the height difference between the highest point and the lowest point in the point cloud set; and estimating the height of the object by removing the ground level from the level difference. Since the ground is relatively low in height, it is empirically known which of the height ranges are ground.
In one embodiment, calculating the updated set of point clouds within the initial annotation frame based on the estimated height of the object may include: points outside the estimated height range of the object are removed from the point cloud set to obtain an updated point cloud set.
In one embodiment, computing an image representation of an object using an updated set of point clouds may include: the updated point cloud set is transformed to an image coordinate system and an image representation of the object is calculated based on the transformed point cloud set. After transforming the point cloud set to the image coordinate system, the point cloud may be converted to an image representation dst using an adaptive threshold function, e.g. opencv. FIG. 2 shows a schematic diagram of an example image representation dst, according to an embodiment of the invention.
Therefore, after the initial labeling frame is obtained, the automatic convergence of the labeling frame can be realized by utilizing the automatic attaching algorithm, so that the labeling efficiency and the labeling accuracy are improved.
After the point cloud data is marked, the final marking frame can be automatically mapped to the 2D image data, so that the point cloud alignment condition is checked on the 2D image in an auxiliary mode.
Because the labeling is completed under the point cloud coordinate system, the points in the labeling frame belong to the BBox coordinate system, and therefore the coordinates of the labeling frame need to be transformed into the 2D image coordinate system through the following transformation:
a) Bbox coordinates-radar coordinates
b) Radar coordinates-camera coordinates
c) Camera coordinates → image coordinates
P I ＝kP C
Wherein, the corner marks L, B, C respectively represent a radar, a point cloud and a camera; p represents coordinates, R represents a rotation matrix, T represents a translation vector, k represents a camera matrix,
rotational moment representing point cloud to radarArray (S)>
Representing a radar-to-camera rotation matrix,
translation vector representing point cloud to radar, +.>
Representing the translation vector of the radar to the camera.
After three transformations, a 4x4 homogeneous coordinate matrix of an image can be obtained as follows:
and (3) performing perspective projection on the calculated matrix into the 2D image by using homogeneous coordinate transformation, namely finishing mapping from the 3D point cloud labeling frame to the pixel position of the 2D image.
In one embodiment, the method 100 may further comprise: and after the continuous frame point cloud data is marked, carrying out multi-frame fitting playback on the marked result. The scene of point cloud and camera dynamic change in the object motion process is simulated through smooth switching between frames, the switching of multiple sensors is supported, the dynamic tracking of the same object annotation at each angle in the annotation process is facilitated, and the annotation precision and efficiency are improved.
The web-based continuous frame point cloud data labeling method can realize multi-frame big data loading, automatically converge labeling frames, improve labeling efficiency, reduce labeling difficulty, enable labeling personnel to finish continuous big data labeling of a specific scene in continuous time, improve data labeling precision and accelerate the production of labeling data; and moreover, the data validity can be checked through multi-frame fitting playback, and the quality barrier of the point cloud data is broken through.
The invention also provides a continuous frame point cloud data labeling device 300 based on the web. Referring to fig. 3, fig. 3 shows a schematic block diagram of a web-based continuous frame point cloud data annotation device 300 according to one embodiment of the invention. As shown in fig. 3, the web-based continuous frame point cloud data annotation device 300 may include a memory 310 and a processor 320, the memory 310 storing a computer program that is executed by the processor 320, which when executed by the processor 320, causes the processor 320 to perform the web-based continuous frame point cloud data annotation method 100 according to the embodiment of the invention described above. Those skilled in the art can understand the specific operation of the web-based continuous frame point cloud data annotation device 300 according to the embodiment of the present invention in combination with the foregoing, and for brevity, the description is omitted herein.
The web-based continuous frame point cloud data labeling device can realize multi-frame big data loading, automatically converge labeling frames, improve labeling efficiency, reduce labeling difficulty, enable labeling personnel to finish continuous big data labeling of a specific scene in continuous time, improve data labeling precision and accelerate the production of labeling data; and moreover, the data validity can be checked through multi-frame fitting playback, and the quality barrier of the point cloud data is broken through.
The present invention also provides a computer readable medium having stored thereon computer executable instructions that, when executed, perform the corresponding steps of the web-based continuous frame point cloud data annotation method 100 described above. Any tangible, non-transitory computer readable medium may be used, including magnetic storage devices (hard disks, floppy disks, etc.), optical storage devices (CD-ROMs, DVDs, blu-ray discs, etc.), flash memory, and/or the like. These computer-executable instructions may be loaded onto a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions which execute on the computer or other programmable data processing apparatus create means for implementing the functions specified. These computer-executable instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including means which implement the function specified. The computer-executable instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified.
Furthermore, according to an embodiment of the present invention, there is also provided a computer program which, when executed by a computer or processor, performs the respective steps of the web-based continuous frame point cloud data annotation method 100 as described above.
Although the illustrative embodiments have been described herein with reference to the accompanying drawings, it is to be understood that the above illustrative embodiments are merely illustrative and are not intended to limit the scope of the present invention thereto. Various changes and modifications may be made therein by one of ordinary skill in the art without departing from the scope and spirit of the invention. All such changes and modifications are intended to be included within the scope of the present invention as set forth in the appended claims.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in order to streamline the invention and aid in understanding one or more of the various inventive aspects, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the description of exemplary embodiments of the invention. However, the method of the present invention should not be construed as reflecting the following intent: i.e., the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this invention.
It will be understood by those skilled in the art that all of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be combined in any combination, except combinations where the features are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments described herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments may be used in any combination.
It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names.
The above description is merely illustrative of specific embodiments of the invention or examples, and the scope of the invention is not limited thereto, but any person skilled in the art will readily appreciate variations or alternatives within the scope of the invention disclosed herein. The protection scope of the invention is subject to the protection scope of the claims.
Claims (11)
1. A web-based continuous frame point cloud data annotation method, the method comprising:
acquiring a labeling request, responding to the labeling request, downloading continuous frame point cloud data to be labeled and corresponding 2D image data, and storing the continuous frame point cloud data and the corresponding 2D image data in a local database;
acquiring an initial annotation frame of an object in one frame of point cloud data in the continuous frame of point cloud data, and automatically calculating based on the initial annotation frame to obtain a final annotation frame; and
the final annotation box is mapped to the 2D image data.
2. The method of claim 1, wherein the method further comprises: and downloading the continuous frame point cloud data and the 2D image data to be annotated and acquiring the initial annotation frame are simultaneously executed by using different processing resources.
3. The method of claim 1, wherein the local database is a local database of a browser.
4. The method of claim 1, wherein automatically computing a final annotation frame based on the initial annotation frame comprises:
acquiring point cloud coordinates of a point cloud set in the initial annotation frame; and
and calculating the minimum circumscribed rectangle of the object based on the point cloud coordinates, and automatically modifying the initial annotation frame according to the minimum circumscribed rectangle to obtain the final annotation frame.
5. The method of claim 4, wherein calculating a minimum bounding rectangle of the object based on the point cloud coordinates comprises:
estimating a height of the object based on the point cloud coordinates, and calculating an updated set of point clouds within the initial annotation frame based on the estimated height of the object;
computing an image representation of the object using the updated set of point clouds; and
and calculating the size of the minimum bounding rectangle based on the image representation of the object, thereby obtaining the minimum bounding rectangle.
6. The method of claim 5, wherein the method further comprises: the point cloud coordinates are transformed from a radar coordinate system to a point cloud coordinate system before estimating the height of the object based on the point cloud coordinates.
7. The method of claim 6, wherein estimating the height of the object based on the point cloud coordinates comprises:
calculating the height difference between the highest point and the lowest point in the point cloud set; and
estimating a height of the object by removing ground level from the level difference, wherein calculating an updated set of point clouds within the initial annotation frame based on the estimated height of the object comprises: points outside the estimated height range of the object are removed from the set of point clouds to obtain an updated set of point clouds.
8. The method of claim 5, wherein computing an image representation of the object using the updated set of point clouds comprises:
transforming the updated set of point clouds to an image coordinate system and calculating an image representation of the object based on the transformed set of point clouds.
9. The method of claim 1, wherein the method further comprises: and after the continuous frame point cloud data is marked, carrying out multi-frame fitting playback on the marked result.
10. A web-based continuous frame point cloud data annotation device, characterized in that the device comprises a memory and a processor, the memory having stored thereon a computer program to be run by the processor, which computer program, when run by the processor, causes the processor to perform the web-based continuous frame point cloud data annotation method according to any of claims 1-9.
11. A computer readable medium having stored thereon computer executable instructions which, when executed, perform the web-based continuous frame point cloud data annotation method of any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211634344.9A CN116245943A (en) | 2022-12-19 | 2022-12-19 | Continuous frame point cloud data labeling method and device based on web |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211634344.9A CN116245943A (en) | 2022-12-19 | 2022-12-19 | Continuous frame point cloud data labeling method and device based on web |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116245943A true CN116245943A (en) | 2023-06-09 |
Family
ID=86625052
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211634344.9A Pending CN116245943A (en) | 2022-12-19 | 2022-12-19 | Continuous frame point cloud data labeling method and device based on web |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116245943A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116665212A (en) * | 2023-07-31 | 2023-08-29 | 福思(杭州)智能科技有限公司 | Data labeling method, device, processing equipment and storage medium |
-
2022
- 2022-12-19 CN CN202211634344.9A patent/CN116245943A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116665212A (en) * | 2023-07-31 | 2023-08-29 | 福思(杭州)智能科技有限公司 | Data labeling method, device, processing equipment and storage medium |
| CN116665212B (en) * | 2023-07-31 | 2023-10-13 | 福思(杭州)智能科技有限公司 | Data labeling method, device, processing equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11954813B2 (en) | Three-dimensional scene constructing method, apparatus and system, and storage medium | |
| KR102145109B1 (en) | Methods and apparatuses for map generation and moving entity localization | |
| US20210142095A1 (en) | Image disparity estimation | |
| CN112613378B (en) | 3D target detection method, system, medium and terminal | |
| CN111402414B (en) | Point cloud map construction method, device, equipment and storage medium | |
| Liang et al. | A survey of 3D object detection | |
| US20210387646A1 (en) | Visual perception method and apparatus, perception network training method and apparatus, device and storage medium | |
| CN110232418B (en) | Semantic recognition method, terminal and computer readable storage medium | |
| CN113536920B (en) | Semi-supervised three-dimensional point cloud target detection method | |
| CN113240734B (en) | Vehicle cross-position judging method, device, equipment and medium based on aerial view | |
| CN116484971A (en) | Automatic driving perception self-learning method and device for vehicle and electronic equipment | |
| CN110599453A (en) | Panel defect detection method and device based on image fusion and equipment terminal | |
| CN116385996B (en) | Multitasking method and device based on three-dimensional matrix camera | |
| Huang et al. | Measuring the absolute distance of a front vehicle from an in-car camera based on monocular vision and instance segmentation | |
| CN116245943A (en) | Continuous frame point cloud data labeling method and device based on web | |
| CN115830265A (en) | Automatic driving movement obstacle segmentation method based on laser radar | |
| CN113989744A (en) | Pedestrian target detection method and system based on oversized high-resolution image | |
| CN110909656B (en) | Pedestrian detection method and system integrating radar and camera | |
| CN115375857A (en) | Three-dimensional scene reconstruction method, device, equipment and storage medium | |
| CN113223146A (en) | Data labeling method and device based on three-dimensional simulation scene and storage medium | |
| CN113436239A (en) | Monocular image three-dimensional target detection method based on depth information estimation | |
| CN117351144A (en) | Dynamic reconstruction method for dam model, terminal equipment and storage medium | |
| CN116012805B (en) | Target perception method, device, computer equipment and storage medium | |
| CN116642490A (en) | Visual positioning navigation method based on hybrid map, robot and storage medium | |
| CN103903269B (en) | The description method and system of ball machine monitor video |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |