CN116204791B - Construction and management method and system for vehicle behavior prediction scene data set - Google Patents
Construction and management method and system for vehicle behavior prediction scene data set Download PDFInfo
- Publication number
- CN116204791B CN116204791B CN202310449345.4A CN202310449345A CN116204791B CN 116204791 B CN116204791 B CN 116204791B CN 202310449345 A CN202310449345 A CN 202310449345A CN 116204791 B CN116204791 B CN 116204791B
- Authority
- CN
- China
- Prior art keywords
- scene
- data
- file
- data set
- obstacle
- 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.)
- Active
Links
- 238000010276 construction Methods 0.000 title claims abstract description 28
- 238000007726 management method Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 23
- 238000007689 inspection Methods 0.000 claims abstract description 11
- 230000033001 locomotion Effects 0.000 claims abstract description 10
- 238000005065 mining Methods 0.000 claims abstract description 10
- 238000007418 data mining Methods 0.000 claims abstract description 8
- 238000001914 filtration Methods 0.000 claims abstract description 6
- 230000006399 behavior Effects 0.000 claims description 43
- 230000011218 segmentation Effects 0.000 claims description 22
- 230000002159 abnormal effect Effects 0.000 claims description 17
- 238000011156 evaluation Methods 0.000 claims description 7
- 238000009826 distribution Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000009499 grossing Methods 0.000 claims description 4
- 238000011002 quantification Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 7
- 238000006243 chemical reaction Methods 0.000 abstract description 2
- 230000009286 beneficial effect Effects 0.000 abstract 1
- 238000012545 processing Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000001133 acceleration Effects 0.000 description 4
- 238000013500 data storage Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012946 outsourcing Methods 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Quality & Reliability (AREA)
- Traffic Control Systems (AREA)
Abstract
The application discloses a construction and management method and a system of a vehicle behavior prediction scene data set, wherein the construction method comprises the steps of collecting driving data of an automatic driving vehicle and surrounding obstacles, segmenting the data, performing data time alignment and format conversion, constructing a scene data file according to a movement interval of the concerned obstacles, repairing the data by using Kalman filtering, and generating a scene tag file describing the scene data; the management method comprises the steps of constructing a scene relation database, a scene playing tool and scene mining to form a subset. The application has the beneficial effects that: the method comprises the steps of constructing a high-quality automatic driving behavior prediction scene data set through massive data acquired by a real vehicle in a real scene, constructing a scene data relation data set, and carrying out data mining by combining a playing tool and a packaged retrieval method, so that the problems of lack of data quality inspection and repair and lack of an efficient management tool in the existing scene data set construction technology are solved.
Description
Technical Field
The application relates to the technical field of vehicle behavior prediction, in particular to a method and a system for constructing and managing a vehicle behavior prediction scene data set.
Background
In recent years, under the hot tide of the study of the automatic driving technology, the vehicle behavior prediction technology is also receiving more and more attention in the industry. Future behavior of vehicles depends on many complex factors, and data-driven predictive algorithms based on deep learning have become increasingly popular. Therefore, the behavior prediction scene data set meeting the algorithm development and verification is reasonably and efficiently constructed to form a key ring of automatic driving behavior prediction research.
The existing automatic driving scene data set construction technology generally comprises the steps of collecting original scene data or outsourcing scene data to form a data source, and then determining information of the automatic driving scene data, wherein the information comprises scene types and data corresponding to the scene types. The process also includes format processing conversion of outsourcing scene data to ensure that all data has the same form. And constructing an automatic driving scene data set by the scene type and the data corresponding to the scene type. However, in the prior art, the original data and the outsourced data are not subjected to checksum correction, which may affect the overall data quality level of the data set, and the management means for the data set is lacking after the data set is constructed into the automatic driving scene data set.
Disclosure of Invention
The application aims to solve the technical problems of the prior art, and provides a method and a system for constructing and managing a vehicle behavior prediction scene data set, which solve the problems of lack of data quality inspection and repair and lack of efficient management tools in the prior scene data set construction technology.
In order to solve the technical problems, the application adopts the following technical scheme:
the construction and management system of the vehicle behavior prediction scene data set comprises a construction system and a management system, wherein the construction system is used for carrying out data acquisition and segmentation, constructing scene tag files and constructing a scene tag database, and the management system is used for carrying out scene mining and playing of corresponding scene files on the basis of the scene tag database.
Further, the building system includes:
the data segmentation module is used for acquiring original data of the automatic driving vehicle and surrounding obstacles, and carrying out data segmentation and alignment based on the original data;
the scene tag file construction module is used for selecting an obstacle set needing special attention based on the data segmentation module, separating scene files according to a non-stationary interval of the obstacle, forming a data set containing all the scene files, and removing abnormal values to generate scene tag files corresponding to each scene file;
the scene data set construction module is used for constructing a relational database based on the scene tag file by taking the file name of the scene file as a main key and the rest information as database fields so as to form a scene tag database.
Further, the management system includes:
the scene mining module is used for searching required scenes through the scene data set management tool to form subsets of different data distributions so as to meet algorithm training and testing requirements under different conditions;
and the playing module is used for playing the corresponding scene file after the scene mining module searches through the database.
The method constructs a high-quality automatic driving behavior prediction scene data set through huge amount of data acquired by a real vehicle in a real scene, constructs a scene data relation data set and performs data mining by combining a playing tool and a packaged searching method.
Further, the method comprises the steps of:
step S1: collecting original data of an automatic driving vehicle and surrounding obstacles, and carrying out data segmentation and alignment based on the original data;
step S2: constructing a scene file, screening out an obstacle set needing special attention according to the behaviors, the distances and the motion states of the obstacles, and separating out the scene file according to the non-stationary interval of the concerned obstacle to form a data set containing all the scene files;
step S3: performing data restoration based on the acquired data set, and generating a scene tag file for each scene data file;
step S4: based on the generated scene tag file, a relational database is established according to the scene information in the scene tag file, the file name of the scene file is used as a main key, and the rest information is used as a database field to form a scene tag database;
step S5: data mining is performed on the scene tag database using scene retrieval to form subsets of different data distributions to meet algorithm training and testing requirements in different situations.
Further, in the step S1, the following rules are simultaneously followed in order when the data are cut:
if the delay of the data of a certain channel between two frames exceeds a preset value, eliminating abnormal data between the two frames and cutting the abnormal data from a delay point to form a plurality of original data segments;
setting a fixed duration threshold interval, segmenting each original data segment exceeding the duration threshold interval, enabling the size of a single data file formed after segmentation to tend to be the average size, enabling the duration corresponding to the single data file to fall in the duration threshold interval, and discarding each original data segment smaller than the duration threshold interval.
Further, in the step S1, the data segmentation performs data alignment according to the receiving time of the obstacle information, frames are extracted from the data with the frame rate higher than the preset value, and linear interpolation frames are performed on the data with the missing frame, so that the frame rate is kept consistent.
Further, in the step S3, the data restoration uses kalman filtering to perform smoothing processing on various data in the scene data file so as to remove abnormal values therein; the plurality of data includes obstacle coordinates, obstacle speeds, and obstacle movement directions in the scene data.
Further, in the step S3, the tag content of the scene tag file includes basic information, behavior information, region information, complexity evaluation and quality inspection items of the scene file; wherein,,
the basic information includes: scene file paths corresponding to the tag files in the scene, scene start-stop time, vehicle information, the number of obstacles in the scene, speed statistics of the obstacles and obstacle type statistics;
the behavior information includes: driving behavior of the vehicle and driving behavior of the obstacle of interest;
the area information includes: driving time periods of the vehicle and the obstacle of interest in different areas;
the complexity evaluation of the scene is quantification of the crowding degree of the scene, the type of the traffic participants and the behavior complexity of the vehicles in the scene;
the quality check item of the scene is a data validity check result of the concerned obstacle in the scene, including whether the speed, the position, the direction and the type have abnormal values or not.
Compared with the prior art, the application has the following technical effects:
the method comprises the steps of constructing a high-quality automatic driving behavior prediction scene data set through massive data acquired by a real vehicle in a real scene, constructing a scene data relation data set, and carrying out data mining by combining a playing tool and a packaged retrieval method, so that the problems of lack of data quality inspection and repair and lack of an efficient management tool in the existing scene data set construction technology are solved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. Like elements or portions are generally identified by like reference numerals throughout the several figures. In the drawings, elements or portions thereof are not necessarily drawn to scale.
FIG. 1 is a step diagram of a construction and management method provided by an embodiment of the present application;
FIG. 2 is a flow chart of data set construction provided by an embodiment of the present application;
FIG. 3 is a schematic diagram of data alignment provided by an embodiment of the present application;
fig. 4 is a schematic diagram of tag file content information provided in an embodiment of the present application;
FIG. 5 is a block diagram of a build system provided by an embodiment of the present application;
fig. 6 is a block diagram of a management system according to an embodiment of the present application.
Detailed Description
Embodiment 1 as shown in fig. 1 and 2, a method for constructing and managing an autopilot vehicle behavior prediction scene data set includes the steps of:
step S1: the method comprises the steps of collecting original data of a vehicle and surrounding obstacles, and carrying out data segmentation and alignment based on the original data.
As a specific example, rosbag format raw data of an autonomous vehicle and surrounding obstacles is collected, including, but not limited to, positioning data of the autonomous vehicle (position coordinates, direction, speed, acceleration, length, width, height, etc.), chassis data of the autonomous vehicle (type, speed, brake value, autopilot status, etc.), perceived data (position coordinates, direction, speed, acceleration, length, width, height, detection frame, type, etc.), and the data is in frame units, and the frame rate of each data is different.
As a specific embodiment, as shown in fig. 3, the data is split, and two splitting rules are followed during the splitting, and processing is firstly performed based on the first rule, rule 1: if there is a data delay in the data, if the delay between two frames of the data of a certain channel is too long, the data needs to be segmented from the delay point to ensure the data to be consistent, the too long delay is considered as an abnormal value, and the length of the delay is set according to an empirical value. Rule 2: setting a fixed duration threshold interval, segmenting each original data segment exceeding the duration threshold interval, enabling the size of a single data file formed after segmentation to tend to be the average size, enabling the duration corresponding to the single data file to fall in the duration threshold interval, and discarding each original data segment smaller than the duration threshold interval.
For example, the preset duration threshold interval is 1800s-7200s, there is a section of 2000s data, there is a section of delay at the end of the data, the next section of 9000s data, there is a section of delay at the end of the data, and the next section of data is 700 s. According to the rule, two sections of delay data intervals are deleted first to form 2000s+9000s+700s original data, a starting point of 2000s, a starting point of 9000s and a starting point of 700s are taken as dividing points, the first section 2000s is completely reserved because of falling in the interval, the second section 9000s is completely reserved because of exceeding the upper limit of the interval, two sections of dividing (the dividing score is obtained by dividing the numerical value of the upper limit of the interval by the total length and rounding upwards) are considered, three sections of dividing can be naturally performed, and the third section 700s is smaller than the lower limit of the interval and is considered to be directly discarded.
As a specific embodiment, since the frame rate of each channel data is different, an alignment operation is required for the subsequent use. The data with higher frame rate is extracted according to the receiving time of the obstacle information, and the missing frame data needs to be linearly interpolated and complemented, so that the final frame rate is kept at 10Hz (10 frames per second).
To facilitate subsequent processing and save data storage space, data is formatted from a Rosbag format to a ProtoBuf format, which is accessed faster than Rosbag and reduces storage space requirements by half.
Step S2: constructing a scene file, screening out an obstacle set needing special attention according to the behaviors, the distances and the motion states of the obstacles, separating out the scene file with the duration of time according to a non-stationary section of the obstacle oi to be noted, and forming a data set containing the scene files, wherein i: representing the number of obstacles in the current scene file that need to be of interest, i=1, 2,3,..n is a total of n obstacles, n: representing the total number of obstacles to be concerned in the current scene file, j: non-stationary interval number j=1, 2,3, representing the current obstacle of interest.The ith obstacle shares +.>Static interval>: representing the total number of stationary intervals of the ith obstacle of interest, < > j->: representing the current obstacle of interest (+)>) Is the j-th stationary interval of (c).
Step S3: based on the acquired data set for data restoration, since the data directly acquired from the vehicle end contains abnormal data caused by positioning jump, perceived quality jump and the like, the Kalman filtering is used for carrying out smoothing processing on various data in the scene data file to remove abnormal values in the scene data file, wherein the abnormal values comprise obstacle coordinates, obstacle speeds and obstacle movement directions in the scene data, and the data after the Kalman filtering is smoother and accords with kinematics.
After the above data is obtained, in order to facilitate the management of the data set at a later stage, a tag file needs to be generated for each scene data file, and tag contents include basic information, behavior information, region information, complexity evaluation of the scene, and quality check items of the scene.
As a specific embodiment, the basic information of the scene file includes a scene file path corresponding to the tag file in the scene, a start-stop time of the scene, automatic driving vehicle information, the number of obstacles in the scene, speed statistics of the obstacles, and obstacle type statistics.
The behavior information includes the driving behavior of the autonomous vehicle and the driving behavior of the obstacle of interest, such as a straight-ahead period, a left-turn period, a lane number where the vehicle is located, and the like.
The area information includes driving periods of the autonomous vehicle and the attention obstacle in different areas, for example, a period of a normal lane, a period of an intersection, and the like.
The complexity evaluation of a scene is a quantification of the congestion level of the scene, the type of traffic participant, and the behavioral complexity of vehicles within the scene.
The quality inspection item of the scene refers to the data validity inspection result of the concerned obstacle in the scene, including whether the speed, the position, the direction, the type and the like have abnormal values.
A schematic diagram of tag file content information as formed in fig. 4.
Step S4: and after the generated scene tag file is based, a relational database is established according to scene information in the tag file, the file name of the scene file is used as a main key, and the rest information is used as a database field. The scene file names of the appointed behaviors, the appointed regions and the appointed motion characteristics can be searched in the scene tag database, then the corresponding files can be taken out for use, and corresponding scene playing tool software which can play the corresponding scene files through the searching result of the database and support the basic playing control function is also arranged. The scene file can be visualized using scene playing software for manual inspection.
Step S5: data mining is performed on a scene tag database by using scene retrieval to form subsets of different data distributions to meet algorithm training and testing requirements under different conditions, such as a scene of focusing on right turn of an obstacle at an intersection, a scene of focusing on the speed of the obstacle to be greater than 20km/h, and a scene containing more than 10 obstacles.
The method constructs a high-quality automatic driving behavior prediction scene data set through huge amount of data acquired by a real vehicle in a real scene, constructs a scene data relation data set, and performs data mining by combining a playing tool and a packaged searching method, thereby solving the problems of lack of data quality inspection and repair and lack of an efficient management tool in the existing scene data set construction technology.
In correspondence to the method for constructing and managing a predicted scenario data set of a driving vehicle behavior of the present application, as shown in fig. 5, a second aspect of an embodiment of the present application provides a system for constructing a predicted scenario data set of a vehicle behavior, including: the system comprises a data segmentation module, a scene tag file construction module and a scene data set construction module. Wherein,,
the data segmentation module collects raw data of the automatic driving vehicle and surrounding obstacles, including but not limited to positioning data (position coordinates, direction, speed, acceleration, length, width, height and the like) of the automatic driving vehicle, chassis data (type, speed, brake value, automatic driving state and the like) of the automatic driving vehicle, sensing data (position coordinates, direction, speed, acceleration, length, width, height, detection frame, type and the like), and the frame rate of each data is different by taking a frame as a unit. The data segmentation module performs data segmentation based on the original data, and sequentially follows two segmentation rules when the data is segmented, including: i) If the delay of the data of a certain channel between two frames exceeds a preset value (exceeds a preset value set according to an empirical value), the abnormal data between the two frames are removed and cut from a delay point; II) setting a fixed time length threshold interval, segmenting each original data segment exceeding the time length threshold interval, enabling the size of a single data file formed after segmentation to be prone to average size, enabling the time length corresponding to the single data file to fall in the time length threshold interval, and discarding each original data segment smaller than the time length threshold interval.
After slicing, since the frame rates of the channel data are different, an alignment operation is required to be performed for subsequent use, the data slicing module aligns according to the receiving time of the obstacle information, frames are extracted from the data with higher frame rate, and the frame data with missing frame data need to be subjected to linear interpolation and frame compensation, so that the final frame rate is kept at 10Hz (10 frames per second). The data is then format converted from Rosbag format to ProtoBuf format, which is accessed faster than Rosbag and reduces the storage space requirement by half, in order to facilitate subsequent processing and save data storage space.
The scene tag file construction module selects an obstacle set needing special attention based on the data segmentation module, separates scene files according to a non-stationary interval of the obstacle, forms a data set containing all the scene files, and then adopts Kalman filtering to carry out smoothing processing on various data (obstacle coordinates, obstacle speed and obstacle movement direction in the scene data) in the scene data files so as to remove abnormal values in the scene data files, and finally generates a data set corresponding to each scene file. On the basis of the above, in order to facilitate the management of the data set in the later stage, a tag file needs to be generated for each scene data file, and the tag content contains basic information, behavior information, region information, complexity evaluation and quality inspection items of the scene file.
The scene data set construction module establishes a relational database based on the scene tag file by taking the file name of the scene file as a main key and the rest information as database fields to form a scene tag database.
A third aspect of the present application provides a management system for a vehicle behavior prediction scene data set, as shown in fig. 6, where a scene mining module and a playing module are set on the basis of a scene tag database constructed by the above construction system, where the scene mining module searches required scenes through a scene data set management tool to form subsets of different data distributions to meet algorithm training and testing requirements under different situations, such as a scene of focusing on a right turn of an obstacle at an intersection, a scene of focusing on an obstacle speed greater than 20km/h, and a scene containing more than 10 obstacles.
As a specific embodiment, the scene mining module establishes a relational database according to scene information in the tag file, takes the file name of the scene file as a primary key, and takes the rest of information as database fields. The scene file names specifying the behavior, region and motion characteristics are retrieved from the scene tag database, and then the corresponding files can be retrieved for use.
And the playing module plays the corresponding scene file through the search result of the database and supports the basic playing control function. The scene file can be visualized using scene playing software for manual inspection.
The preferred embodiments of the present application have been described in detail above, but the present application is not limited to the specific details of the above embodiments, and various equivalent changes (such as number, shape, position, etc.) may be made to the technical solution of the present application within the scope of the technical concept of the present application, and these equivalent changes all belong to the protection of the present application.
Claims (5)
1. The construction and management method of the vehicle behavior prediction scene data set is characterized by comprising the following steps of: the method is applied to a construction and management system of the vehicle behavior prediction scene data set, the system comprises a construction system and a management system, the construction system is used for carrying out data acquisition and segmentation, constructing a scene tag file and constructing a scene tag database, and the management system is used for carrying out scene mining and playing of a corresponding scene file on the basis of the scene tag database;
the method comprises the steps of constructing a high-quality automatic driving behavior prediction scene data set through huge amount of data acquired by a real vehicle in a real scene, constructing a scene data relation data set, and carrying out data mining by combining a playing tool and a packaged searching method;
the method comprises the following steps:
step S1: collecting original data of an automatic driving vehicle and surrounding obstacles, and carrying out data segmentation and alignment based on the original data;
step S2: constructing a scene file, screening out an obstacle set needing special attention according to the behaviors, the distances and the motion states of the obstacles, and separating out the scene file according to the non-stationary interval of the concerned obstacle to form a data set containing all the scene files;
step S3: performing data restoration based on the acquired data set, and generating a scene tag file for each scene data file;
step S4: based on the generated scene tag file, a relational database is established according to the scene information in the scene tag file, the file name of the scene file is used as a main key, and the rest information is used as a database field to form a scene tag database;
step S5: performing data mining on a scene tag database by using scene retrieval to form subsets of different data distributions to meet algorithm training and testing requirements under different conditions;
in the step S1, the following rules are simultaneously followed according to the sequence when the data are cut:
if the delay of the data of a certain channel between two frames exceeds a preset value, eliminating abnormal data between the two frames and cutting the abnormal data from a delay point to form a plurality of original data segments;
setting a fixed time length threshold interval, segmenting each original data segment exceeding the time length threshold interval, enabling the size of a single data file formed after segmentation to tend to be the average size, enabling the time length corresponding to the single data file to fall in the time length threshold interval, and discarding each original data segment smaller than the time length threshold interval;
and in the step S1, data alignment is carried out according to the receiving time of the obstacle information, frame extraction is carried out on data with the frame rate higher than a preset value, and linear interpolation frame compensation is carried out on data with the missing frame so as to keep the frame rate consistent.
2. The method of constructing and managing a vehicle behavior prediction scene data set according to claim 1, characterized in that: in the step S3, the data restoration adopts Kalman filtering to carry out smoothing treatment on various data in the scene data file so as to remove abnormal values in the scene data file; the plurality of data includes obstacle coordinates, obstacle speeds, and obstacle movement directions in the scene data.
3. The method of constructing and managing a vehicle behavior prediction scene data set according to claim 1, characterized in that: the label content of the scene label file in the step S3 comprises basic information, behavior information, region information, complexity evaluation and quality inspection items of the scene file; wherein,,
the basic information includes: scene file paths corresponding to the tag files in the scene, scene start-stop time, vehicle information, the number of obstacles in the scene, speed statistics of the obstacles and obstacle type statistics;
the behavior information includes: driving behavior of the vehicle and driving behavior of the obstacle of interest;
the area information includes: driving time periods of the vehicle and the obstacle of interest in different areas;
the complexity evaluation of the scene is quantification of the crowding degree of the scene, the type of the traffic participants and the behavior complexity of the vehicles in the scene;
the quality check item of the scene is a data validity check result of the concerned obstacle in the scene, including whether the speed, the position, the direction and the type have abnormal values or not.
4. The method of constructing and managing a vehicle behavior prediction scene data set according to claim 1, characterized in that: the construction system comprises:
the data segmentation module is used for acquiring original data of the automatic driving vehicle and surrounding obstacles, and carrying out data segmentation and alignment based on the original data;
the scene tag file construction module is used for selecting an obstacle set needing special attention based on the data segmentation module, separating scene files according to a non-stationary interval of the obstacle, forming a data set containing all the scene files, and removing abnormal values to generate scene tag files corresponding to each scene file;
the scene data set construction module is used for constructing a relational database based on the scene tag file by taking the file name of the scene file as a main key and the rest information as database fields so as to form a scene tag database.
5. The method of constructing and managing a vehicle behavior prediction scene data set according to claim 1, characterized in that: the management system includes:
the scene mining module is used for searching required scenes through the scene data set management tool to form subsets of different data distributions so as to meet algorithm training and testing requirements under different conditions;
and the playing module is used for playing the corresponding scene file after the scene mining module searches through the database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310449345.4A CN116204791B (en) | 2023-04-25 | 2023-04-25 | Construction and management method and system for vehicle behavior prediction scene data set |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310449345.4A CN116204791B (en) | 2023-04-25 | 2023-04-25 | Construction and management method and system for vehicle behavior prediction scene data set |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116204791A CN116204791A (en) | 2023-06-02 |
| CN116204791B true CN116204791B (en) | 2023-08-11 |
Family
ID=86515973
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310449345.4A Active CN116204791B (en) | 2023-04-25 | 2023-04-25 | Construction and management method and system for vehicle behavior prediction scene data set |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116204791B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117681893B (en) * | 2024-02-02 | 2024-05-14 | 中国科学院自动化研究所 | Mining area scene-oriented automatic driving data construction method and device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107235044A (en) * | 2017-05-31 | 2017-10-10 | 北京航空航天大学 | It is a kind of to be realized based on many sensing datas to road traffic scene and the restoring method of driver driving behavior |
| DE102017009971A1 (en) * | 2017-10-26 | 2019-05-02 | Daimler Ag | A method of testing a lane keeping assistance system for a vehicle |
| CN110083163A (en) * | 2019-05-20 | 2019-08-02 | 三亚学院 | A kind of 5G C-V2X bus or train route cloud cooperation perceptive method and system for autonomous driving vehicle |
| CN110691990A (en) * | 2017-06-02 | 2020-01-14 | 华为技术有限公司 | Obstacle detection method and equipment |
| CN111199087A (en) * | 2018-10-31 | 2020-05-26 | 百度在线网络技术（北京）有限公司 | Scene recognition method and device |
| CN114676844A (en) * | 2020-12-24 | 2022-06-28 | 北京百度网讯科技有限公司 | Method, apparatus, device, computer-readable storage medium for automatic driving |
| CN115718702A (en) * | 2022-11-23 | 2023-02-28 | 武汉中海庭数据技术有限公司 | Automatic driving test scene library construction method and system |
| DE102022112060B3 (en) * | 2022-05-13 | 2023-04-20 | Dr. Ing. H.C. F. Porsche Aktiengesellschaft | Scenario database for a method and a system for calibrating and validating a driver assistance system (ADAS) and/or an automated driving system (ADS) |
-
2023
- 2023-04-25 CN CN202310449345.4A patent/CN116204791B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107235044A (en) * | 2017-05-31 | 2017-10-10 | 北京航空航天大学 | It is a kind of to be realized based on many sensing datas to road traffic scene and the restoring method of driver driving behavior |
| CN110691990A (en) * | 2017-06-02 | 2020-01-14 | 华为技术有限公司 | Obstacle detection method and equipment |
| DE102017009971A1 (en) * | 2017-10-26 | 2019-05-02 | Daimler Ag | A method of testing a lane keeping assistance system for a vehicle |
| CN111199087A (en) * | 2018-10-31 | 2020-05-26 | 百度在线网络技术（北京）有限公司 | Scene recognition method and device |
| CN110083163A (en) * | 2019-05-20 | 2019-08-02 | 三亚学院 | A kind of 5G C-V2X bus or train route cloud cooperation perceptive method and system for autonomous driving vehicle |
| CN114676844A (en) * | 2020-12-24 | 2022-06-28 | 北京百度网讯科技有限公司 | Method, apparatus, device, computer-readable storage medium for automatic driving |
| DE102022112060B3 (en) * | 2022-05-13 | 2023-04-20 | Dr. Ing. H.C. F. Porsche Aktiengesellschaft | Scenario database for a method and a system for calibrating and validating a driver assistance system (ADAS) and/or an automated driving system (ADS) |
| CN115718702A (en) * | 2022-11-23 | 2023-02-28 | 武汉中海庭数据技术有限公司 | Automatic driving test scene library construction method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116204791A (en) | 2023-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110634291B (en) | High-precision map topology automatic construction method and system based on crowdsourcing data | |
| KR102309711B1 (en) | Cnn-based learning method, learning device for selecting useful training data and test method, test device using the same | |
| CN106570453B (en) | Method, device and system for pedestrian detection | |
| CN110751224A (en) | Training method of video classification model, video classification method, device and equipment | |
| CN116204791B (en) | Construction and management method and system for vehicle behavior prediction scene data set | |
| CN111796917B (en) | Operator operation scheduling method and device | |
| WO2021185121A1 (en) | Model generation method and apparatus, object detection method and apparatus, device, and storage medium | |
| CN112965960A (en) | Wisdom police service data fusion washs and studies out and judges device | |
| CN112905849A (en) | Vehicle data processing method and device | |
| CN107992937A (en) | Unstructured data decision method and device based on deep learning | |
| CN112269827B (en) | Data processing method and device, computer equipment and computer readable storage medium | |
| CN111582214A (en) | Twin network-based behavior analysis method, system and device for cage-raised animals | |
| CN112466118A (en) | Vehicle driving behavior recognition method, system, electronic device and storage medium | |
| CN114926791A (en) | Method and device for detecting abnormal lane change of vehicles at intersection, storage medium and electronic equipment | |
| CN115545281A (en) | Risk scene identification method, device, equipment and storage medium | |
| CN112329152B (en) | Driving data searching method, device and equipment | |
| CN117093260B (en) | Fusion model website structure analysis method based on decision tree classification algorithm | |
| Ding et al. | Mit-avt clustered driving scene dataset: Evaluating perception systems in real-world naturalistic driving scenarios | |
| CN117194698B (en) | Task processing system and method based on OAR semantic knowledge base | |
| CN114332797A (en) | Road scene semantic segmentation method and system with self-evaluation mechanism | |
| CN115027497B (en) | Target vehicle cut-in intention prediction method and readable storage medium | |
| CN116740495A (en) | Training method and defect detection method for defect detection model of road and bridge tunnel | |
| CN115422178A (en) | Vehicle data cleaning method and device, electronic equipment and storage medium | |
| CN113963204A (en) | Twin network target tracking system and method | |
| CN112277958B (en) | Driver braking behavior analysis method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |