CN113537046A - Map lane marking method and system based on vehicle track big data detection - Google Patents
Map lane marking method and system based on vehicle track big data detection Download PDFInfo
- Publication number
- CN113537046A CN113537046A CN202110795387.4A CN202110795387A CN113537046A CN 113537046 A CN113537046 A CN 113537046A CN 202110795387 A CN202110795387 A CN 202110795387A CN 113537046 A CN113537046 A CN 113537046A
- Authority
- CN
- China
- Prior art keywords
- lane
- vehicle
- data
- lane line
- map
- 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.)
- Withdrawn
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/05—Geographic models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
- G06T7/344—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/521—Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light
-
- 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
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30204—Marker
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Computer Graphics (AREA)
- Optics & Photonics (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a map lane marking method and a map lane marking system based on vehicle track big data detection, which comprise the following steps: step S1: data acquisition is carried out on the verified road section through acquisition equipment; step S2: acquiring driving track information according to the vehicle model; step S3: establishing a mapping relation between a lane line and a driving track in the marked road section and a lane line and a driving track in the verified road section; step S4: establishing a relation model of a lane line and a lane track of a verification road section and correcting; step S5: and predicting the lane lines of other road sections through a relation model of the lane lines and the lane tracks, and finishing the lane line marking of the map. Compared with the traditional lane line marking method, the lane track data are processed, so that a small amount of error data cannot influence the correct lane line, the lane line marking accuracy is improved, and the lane line gradually converges towards the real lane line along with the increase of the data volume, so that the lane line marking method has stronger robustness.
Description
Technical Field
The invention relates to a map lane line marking method, in particular to a map lane line marking method and system based on vehicle track big data detection.
Background
The existing map lane marking technology mainly comprises the following steps:
1. the image set is collected through a camera, lane line information in the image set is identified through deep learning, and the lane line information of the high-precision map is spliced by combining pose information of vehicles. In the method, the vision is seriously influenced by light, raining, accumulated water on the road surface and damage and shielding of the lane line can all influence the recognition result, and the accuracy of the high-accuracy map is difficult to break through 10cm grade because the camera cannot provide depth information, and a few pure vision high-accuracy map lane line marking schemes which can reach centimeter grade are adopted.
2. The method comprises the steps of collecting lane line information through a laser radar, combining pose information of a vehicle, splicing a point cloud map, identifying a corresponding lane line through deep learning of generated point cloud, and combining manual marking to generate a high-precision map.
The laser radar can not provide color information of a lane line, white lines and yellow lines can not be distinguished, the lane line point cloud identified by the laser radar is not obviously different from other point cloud positions, false detection is easy to occur, the problem of missed detection is solved, the laser radar is easy to interfere, rain falls, accumulated water on the road surface, other laser radars are shot oppositely, the lane line is damaged, the problem of shielding and the like is solved, and the identification result can be influenced.
In the chinese invention patent application with publication number CN112329553A, a lane marking method and apparatus are disclosed, the method comprising: acquiring an image to be marked, wherein the image to be marked comprises at least one lane line; determining a tangent circle which is tangent to the boundaries of the two sides of each lane line at the two ends of each lane line respectively; acquiring lane line fitting parameters, and performing curve fitting based on the circle center of the tangent circle and the lane line fitting parameters to obtain a lane center line of the lane line; determining lane boundary lines of boundaries at two sides of the lane line according to the tangent circle and the lane center line; and marking the lane line according to the lane boundary line. According to the marking method, the lane central line of the lane line is firstly fitted by utilizing the tangent circle, and then the lane central line is expanded to obtain the lane boundary lines of the boundaries at the two sides of the lane line, so that the marking efficiency of the lane line can be improved, and the marking width of the lane line can be ensured to be changed uniformly.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a map lane marking method and system based on large data of detected vehicle tracks.
The invention provides a map lane marking method based on vehicle track big data detection, which comprises the following steps:
step S1: acquiring data of a verified road section through acquisition equipment, wherein the verified road section is a road section which is not marked with a lane line on a map;
step S2: acquiring driving track information according to the vehicle model;
step S3: establishing a mapping relation between a lane line and a driving track in a marked road section and a lane line and a driving track in a verified road section, wherein the marked road section is a road section marked with the lane line on a map;
step S4: establishing a relation model of a lane line and a lane track of a verification road section and correcting;
step S5: and predicting the lane lines of other road sections through a relation model of the lane lines and the lane tracks, and finishing the lane line marking of the map.
Preferably, in step S3, the map of the city is rasterized, and the vehicle track points are converted into a grid probability density corresponding to the lane and the road center line, so as to obtain a maximum likelihood lane line distribution.
Preferably, in step S4, a relationship model between lane lines and lane tracks in a high-precision map area may be obtained by performing logical modeling or end-to-end machine learning training on a mapping relationship between lane lines and lane tracks, and a lane line in an area where the verification data set is located is predicted; meanwhile, the predicted lane line is compared with the real lane line, and the model parameters are continuously corrected.
Preferably, in step S2, the data collected by the sensor is calculated to obtain the position of the vehicle in the image relative coordinate system, then the envelope frame of the vehicle object collected by the sensor in the three-dimensional coordinate system is estimated, or the image with depth information is used to perform vehicle identification, the identified vehicle is modeled, and the trajectory of the vehicle object in the world coordinate system is calculated through the established vehicle model and the position relationship with the vehicle.
Preferably, for vehicle identification of different vehicle types, a vehicle registration model trained in advance is obtained through training of different vehicle models.
The invention provides a map lane marking system based on vehicle track big data detection, which comprises: the data acquisition module and the cloud processing module are used for processing and analyzing the data acquired by the data acquisition module to form lane line data of the road section.
Preferably, the data acquisition module comprises the following modules:
a motion module: providing exercise capacity for the acquisition equipment;
a fusion positioning module: an independent or fusion positioning mode is adopted to provide real-time positioning information for acquisition equipment;
a sensor module: the system is used for acquiring the running track information of vehicles in the sensing range of the acquisition equipment on the road;
the control and processing module: the system is used for controlling the motion of the acquisition equipment, collecting and storing sensor data and calculating and storing the vehicle track in the range of the sensor.
Preferably, the data acquisition module further comprises a data transmission module, and the data transmission module transmits the information acquired by the acquisition equipment and each sensor to the information processing module.
Preferably, the cloud processing module includes the following modules:
a data processing module: collecting data of various sensors, and forming a data set under a unified time axis by one or more sensor data through time sequence synchronization;
a vehicle trajectory extraction module: extracting a vehicle track under a world coordinate axis by combining data of each sensor;
model training and regression module: training the obtained vehicle track and the lane line of a known map, so as to predict the lane line through the lane track;
lane marking module: using the trained model for marking the lane line of the unmarked area;
among the above modules, the latter module depends on the data of the former module.
Preferably, the sensor module comprises a vision sensor, a laser radar sensor or other sensors capable of providing depth, image and point cloud information, and the vision sensor and the laser radar sensor are synchronously sampled through GNSS or hardware; when the laser radar sensor works, the key frame point cloud with dense point cloud density of one frame is obtained by splicing multi-frame laser radar data.
Compared with the prior art, the invention has the following beneficial effects:
1. and tracking the motion track of the vehicles around the acquisition equipment, judging the position of the lane line through a large data set for tracking the vehicle track, and obtaining a real lane line through data convergence of a large number of vehicle tracks.
2. Compared with the traditional lane marking method, the lane trajectory data are processed, so that a small amount of error data cannot influence the correct lane, and the lane marking accuracy is improved.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is an architectural design diagram of a map lane marking method based on vehicle track big data detection according to the present application;
FIG. 2 is a vehicle track point flow chart of a map lane marking method based on vehicle track big data detection according to the present application;
FIG. 3 is a schematic diagram of a probability density fitting process of a map lane marking method based on big data of detected vehicle tracks according to the present application;
FIG. 4 is a projection diagram of a vehicle track point under a map grid in a map lane marking method based on vehicle track big data detection according to the present application;
fig. 5 is a probability distribution diagram of a grid around which a track point falls on a grid map to become a vehicle center line in the map lane marking method based on vehicle track big data detection according to the present application;
fig. 6 is a probability distribution of a grid around a grid map into a lane line in a map lane line labeling method based on vehicle track big data detection according to the present application.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
Referring to fig. 1, a map lane marking system based on vehicle track big data detection comprises a data acquisition module and a cloud processing module, wherein one or more data acquisition modules transmit collected data back to the cloud processing module through the collected data, and the cloud processing module processes and calculates the collected data to obtain map data marked with a lane.
Wherein, the data acquisition module includes four at least modules: motion module fuses orientation module, sensor module, control and processing module, and under most conditions, still include data transmission module to the realization is to real-time data's high in the clouds transmission and remote instruction's receive function, but this module is not necessary, and in principle, off-line data collection realizes through hardware transmission that calculation also should be listed in the scope of protection of this patent.
The motion module can support manual driving, operation, also can support autopilot, remote control, and its effect is for providing the motion ability for data acquisition hardware, can accomplish the collection of sensor data at ordinary collection highway section, and the motion module commonly used is the power module of vehicle, or unmanned aerial vehicle's flight module to this module can be used in the road and travel.
The fusion positioning module can provide real-time positioning information for the acquisition equipment by fusing or independently adopting positioning data provided by visual positioning, GPS or other types of satellite positioning, laser radar point cloud positioning, wheel speed, IMU and other sensors which can be used for direct or auxiliary positioning, and the data provided by the module can be understood as providing relatively accurate overall position with time sequence in the road driving process or local position capable of being converted into overall position for the acquisition equipment.
The sensor module should include current sensor modules such as monocular camera, multi-view camera, depth camera to and single line laser radar, multi-line laser radar, solid state laser radar, unmanned driving such as millimeter wave radar, the distance that unmanned aerial vehicle and automatic robot field are commonly used, image and point cloud sensor, should still include the sensing equipment that is used for discerning vehicle and road in the trade of other kinds that appear in the future. The module is used for acquiring the driving track of the vehicle in the sensing range of the acquisition equipment on the road or obtaining direct or auxiliary information of the driving track, such as vehicle images, vehicle point clouds or simplified position information of other vehicle models. It should be noted that if a communication structure capable of directly transmitting the position information of the running vehicle appears in the future, such as a related device directly acquiring the position information of the vehicle close to the road through bluetooth or other 5G devices, the communication structure also belongs to the description category of the sensor and the protection field of the scheme.
The control and processing module comprises but is not limited to an existing unmanned vehicle control module, an FPGA, a single chip microcomputer, an industrial personal computer, an ARM board card, or a CPU and a GPU which have image processing or point cloud processing and storage functions or software and hardware equipment which have the same or can play the same effect in function. The functions of the method may include controlling the motion of the acquisition device, collecting and storing sensor data, and calculating and storing vehicle trajectories within the sensor range, it should be noted that whether the vehicle trajectories are calculated in real time is not within the protection scope of the method, and the behavior of direct or auxiliary lane marking, whether online or offline, should be understood as extending within the scope set forth in the method as long as the sensor data capable of generating the vehicle trajectories are collected.
If necessary, the method should support a data transmission module, and after the communication module is mounted, the sensor or the vehicle pose and the sensor information are transmitted to a remote vehicle track generation module in any mode under any communication protocol, and the behavior of marking the lane line is understood to be within the range stated by the method.
The cloud processing module comprises a data processing module, a vehicle track extraction module, a model training and regression module and a lane marking module. The latter module relies on the data of the former module, and from the end-to-end angle, the process from a large amount of acquisition equipment sensors and pose information to the marked lane line may need to be corrected by human continuously, but no matter the quality of the generated lane line is good or bad, the lane line is within the protection range of the method as long as the track information of the road vehicle is used.
And the data processing module is used for collecting various sensor data of the vehicle sensors and forming one or more sensor data into a data set under a unified time axis through time sequence synchronization.
The vehicle track extraction module extracts a vehicle track by combining data of each sensor, wherein the vehicle track is in a world coordinate system, so that all the equipment acquisition equipment can be used universally, and the extracted vehicle track is within the protection range of the method whether containing or not containing timestamp information, namely the vehicle track required to be protected by the method can only contain spatial information.
The model training and regression module trains the obtained vehicle track and the lane line of the known high-precision map, and the model training and regression module is used for training the obtained vehicle track and the lane line of the known high-precision map, wherein the training can be a deep learning method or a traditional regression analysis method, namely, the prediction of the lane line is generated only through the track, namely the protection range of the scheme is obtained, and other auxiliary data such as lane line information extracted by a sensor or manually marked lanes and the like can be adopted during the period without influencing the protection range of the scheme.
And the lane marking module is used for marking the trained model on the lane without the pre-marked area, may include an automatic part and a manual correction part, and is commonly used in various other high-precision lane marking methods, and is not described herein again.
A map lane line marking method based on vehicle track big data detection comprises the following steps:
step S1: and carrying out data acquisition on the verified road section through acquisition equipment. The method comprises the steps of acquiring images, point clouds or other forms of road surfaces and other data information on lanes in a sensor acquisition range through a movable acquisition device and a sensor which is carried by the device and is rigidly fixed with the device, and acquiring pose information corresponding to the information time sequence of the sensor through a single positioning device of the device or a positioning module embedded into a corresponding positioning algorithm.
The acquisition equipment comprises a sensor for shooting road surface information, and comprises a vision sensor and a laser radar sensor, wherein the adopted vision sensor is a monocular camera, the adopted laser radar sensor is a 32 or 64-line laser radar, firstly, the laser radar and a camera sensor timestamp are synchronized to the same time axis through GNSS or hardware synchronization, 5HZ is uniformly adopted as sampling frequency, and at the moment, the laser radar and image data which are matched one by one and have the frequency of 5HZ can be obtained.
Step S2: and acquiring a driving track according to the vehicle model. The vehicle object in the sensor information is identified through the existing deep learning algorithm, modeling is carried out, the coordinates of the vehicle model identified by each frame in the world coordinate system are calculated through multi-frame sensor data and the current position and posture of the acquisition equipment, and the track information of the vehicle model is obtained through splicing.
Step S3: and establishing a mapping relation between the lane lines and the driving tracks, and predicting the lane tracks. According to traffic laws and regulations and common driving habits, a certain geometric relationship exists between the lane lines and a large number of driving tracks, and a mapping relationship can be finally obtained between the lane lines of the road section of the existing high-precision map and the large number of vehicle tracks of the road section by means of deep learning or traditional regression analysis.
Step S4: and establishing a relation model of the lane line and the lane track and correcting. Through modeling analysis of the mapping relation or end-to-end deep learning, a relation model of a lane line and a lane track of an area where a high-precision map is not established can be obtained, the lane line of the area where the verification data set is located is predicted through the obtained model, and the model parameters are corrected by comparing the predicted lane line with a real lane line.
Step S5: and predicting the lane line of the unknown area through the obtained relationship model of the lane line and the lane track, and finishing the lane line marking of the map.
Referring to fig. 2, the data sources used in the method of the present invention are sensor data sets and acquisition device pose data sets acquired by a mobile device (such as a data acquisition vehicle, a drone or other human-driven device or automatic device that can carry sensors to realize movement) carrying a laser radar and a camera or other sensors including but not limited to images, point clouds and other sensors reflecting the position information of vehicles and lane lines in roads.
After sampling, inputting the image into yolo or other trained deep learning models, identifying vehicle information in an original image acquired in the operation process of acquisition equipment, calculating data acquired by a sensor to obtain the position of an acquired vehicle in an image relative coordinate system, then estimating an envelope frame of a vehicle object acquired by the sensor in a three-dimensional coordinate system, or using the image with the depth information to identify the vehicle, modeling the identified vehicle, and calculating the track of the vehicle object in a world coordinate system through the established vehicle model and the position relation of the vehicle object and the acquired vehicle.
The single-frame laser radar data often lack complete vehicle point clouds, and a key frame point cloud corresponding to an image with dense point cloud density is obtained usually in a mode of splicing 3-5 frames in space. Through the CNN deep learning network, the image prior information is used as a coarse screen, or the original point cloud information is directly used for matching, so that the point cloud of the vehicle object in the radar view field can be marked in the original point cloud frame.
For vehicles of different vehicle types, a vehicle registration model trained in advance can be obtained through training of different vehicle models, meanwhile, a simple scheme can be adopted in the step, such as matching of front wheels and rear wheels of the vehicles or other vehicle body components is performed, vehicle point cloud can be completely mapped to a vehicle point cloud model in a world coordinate system through registration, vehicle track points are extracted according to the point cloud model, generally, entity part points of the vehicles, which are in contact with the ground, or projection points of a certain specific point position of the vehicles on the ground are selected as track point references, and different vehicle types can select one point in a surface curve in the world coordinate system in a unified mode as a reference point of the vehicle track.
The lane lines generated by the method are established on two basic assumptions, 1, the driving track of the travel when the driver drives and the lane lines statistically meet a certain geometric relationship, namely, most of the drivers should distribute vehicles in the center of the lane as much as possible in the process of driving the vehicles to go straight, and the left and right deviation should meet or approximately meet normal distribution on the premise of no other precondition. 2. When the data is sufficient, exceptional conditions caused by driver's wrong or irregular driving, traffic accidents or other external conditions will not have significant influence on the result because the data volume is too small and the proportion is too low, that is, the lane marking result is continuously converged to the real lane due to the increase of the data volume.
Referring to fig. 3, a rasterization method may be used, and first, a high-precision map of a city is rasterized, for example, the map is divided into map grids of 100m × 100m, and all tracks falling within the grids at this time may be understood as an object of single regression analysis, as shown in fig. 4; the grid is regridded into 5cm by 5cm probability grids, and at this time, the probability that each grid is divided into lane lines, road center lines or other types of icon lines with high precision only needs to be calculated.
Referring to fig. 5 and 6, it can be known from experience that, for a trace point on each trace, a probability grid near the point is considered to be a probability distribution of the point on the road centerline, and the closer to the trace point, the higher the probability of becoming the road center point is, so we can design a probability updating function for each grid to ensure that when a new trace point falls within the grid, the probability of the grids around the point becoming the road centerline is updated accordingly. Similarly, in the grids which are 1 and 2 away from the lane line and parallel to the driving direction of the vehicle, the highest probability is the candidate point of the lane line, which is also based on the fact that the driver tends to drive along the center direction of the lane under big data, so far, for each vehicle track point, the probability grid probability density of the corresponding lane and the center line of the lane can be updated, the training process of the model can adopt the traditional linear regression analysis, and can also adopt the deep learning mode to extract the residual error with the real lane line for optimization, and finally the maximum likelihood lane line distribution is obtained.
Those skilled in the art will appreciate that, in addition to implementing the system and its various devices, modules, units provided by the present invention as pure computer readable program code, the system and its various devices, modules, units provided by the present invention can be fully implemented by logically programming method steps in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices, modules and units thereof provided by the invention can be regarded as a hardware component, and the devices, modules and units included in the system for realizing various functions can also be regarded as structures in the hardware component; means, modules, units for performing the various functions may also be regarded as structures within both software modules and hardware components for performing the method.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (10)
1. A map lane line marking method based on vehicle track big data detection is characterized by comprising the following steps:
step S1: acquiring data of a verified road section through acquisition equipment, wherein the verified road section is a road section which is not marked with a lane line on a map;
step S2: acquiring driving track information according to the vehicle model;
step S3: establishing a mapping relation between a lane line and a driving track in a marked road section and a lane line and a driving track in a verified road section, wherein the marked road section is a road section marked with the lane line on a map;
step S4: establishing a relation model of a lane line and a lane track of a verification road section and correcting;
step S5: and predicting the lane lines of other road sections through a relation model of the lane lines and the lane tracks, and finishing the lane line marking of the map.
2. The map lane marking method based on the monitoring vehicle track big data as claimed in claim 1, wherein: in step S3, the map of the city is rasterized, and the vehicle track points are converted into grid probability densities of corresponding lanes and road center lines, so as to obtain a lane line distribution with the maximum likelihood.
3. The method for building a relationship model between a lane line and a lane track according to claim 1, wherein: in step S4, a relationship model between lane lines and lane tracks in a high-precision map area may be obtained by performing logical modeling on a mapping relationship between lane lines and lane tracks or performing end-to-end machine learning training, so as to predict a lane line in an area where the verification data set is located; meanwhile, the predicted lane line is compared with the real lane line, and the model parameters are continuously corrected.
4. The method for building a relationship model between a lane line and a lane track according to claim 1, wherein: in step S2, the data acquired by the sensor is calculated to obtain the position of the vehicle in the image relative coordinate system, and then the envelope frame of the vehicle object acquired by the sensor in the three-dimensional coordinate system is estimated, or the image with depth information is used to identify the vehicle, model building is performed on the identified vehicle, and the trajectory of the vehicle object in the world coordinate system is calculated through the established vehicle model and the position relationship with the vehicle.
5. The map lane marking method based on the big data of the monitored vehicle track according to claim 4, wherein the method comprises the following steps: and for vehicle identification of different vehicle types, training different vehicle models to obtain a vehicle registration model trained in advance.
6. The utility model provides a map lane line mark system based on detect vehicle orbit big data which characterized in that: the method comprises the following steps: the data acquisition module and the cloud processing module are used for processing and analyzing the data acquired by the data acquisition module to form lane line data of the road section.
7. The map lane marking system based on the big data of the monitored vehicle track according to claim 6, wherein: the data acquisition module comprises the following modules:
a motion module: providing exercise capacity for the acquisition equipment;
a fusion positioning module: an independent or fusion positioning mode is adopted to provide real-time positioning information for acquisition equipment;
a sensor module: the system is used for acquiring the running track information of vehicles in the sensing range of the acquisition equipment on the road;
the control and processing module: the system is used for controlling the motion of the acquisition equipment, collecting and storing sensor data and calculating and storing the vehicle track in the range of the sensor.
8. The map lane marking system based on the big data of the monitored vehicle track according to claim 7, wherein: the data acquisition module further comprises a data transmission module, and the data transmission module transmits the information acquired by the acquisition equipment and the sensors to the information processing module.
9. The map lane marking system based on the big data of the monitored vehicle track according to claim 6, wherein: the cloud processing module comprises the following modules:
a data processing module: collecting data of various sensors, and forming a data set under a unified time axis by one or more sensor data through time sequence synchronization;
a vehicle trajectory extraction module: extracting a vehicle track under a world coordinate axis by combining data of each sensor;
model training and regression module: training the obtained vehicle track and the lane line of a known map, so as to predict the lane line through the lane track;
lane marking module: using the trained model for marking the lane line of the unmarked area;
among the above modules, the latter module depends on the data of the former module.
10. The map lane marking system based on the big data of the monitored vehicle track according to claim 7, wherein: the sensor module comprises a vision sensor, a laser radar sensor or other sensors capable of providing depth, image and point cloud information, and the vision sensor and the laser radar sensor synchronously sample through GNSS or hardware; when the laser radar sensor works, the key frame point cloud with dense point cloud density of one frame is obtained by splicing multi-frame laser radar data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110795387.4A CN113537046A (en) | 2021-07-14 | 2021-07-14 | Map lane marking method and system based on vehicle track big data detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110795387.4A CN113537046A (en) | 2021-07-14 | 2021-07-14 | Map lane marking method and system based on vehicle track big data detection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113537046A true CN113537046A (en) | 2021-10-22 |
Family
ID=78099054
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110795387.4A Withdrawn CN113537046A (en) | 2021-07-14 | 2021-07-14 | Map lane marking method and system based on vehicle track big data detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537046A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114119671A (en) * | 2021-12-01 | 2022-03-01 | 清华大学 | Multi-target tracking method based on occlusion compensation and used for three-dimensional space information fusion |
| CN114373298A (en) * | 2021-12-16 | 2022-04-19 | 苏州思卡信息系统有限公司 | Method for calculating lane number by adopting roadside laser radar |
| CN114494618A (en) * | 2021-12-30 | 2022-05-13 | 广州小鹏自动驾驶科技有限公司 | Map generation method and device, electronic equipment and storage medium |
| CN115937812A (en) * | 2023-01-06 | 2023-04-07 | 河北博士林科技开发有限公司 | Method and system for generating virtual lane line at traffic intersection |
| CN116679293A (en) * | 2023-08-01 | 2023-09-01 | 长沙隼眼软件科技有限公司 | Multi-radar target track splicing method and device based on high-precision map |
-
2021
- 2021-07-14 CN CN202110795387.4A patent/CN113537046A/en not_active Withdrawn
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114119671A (en) * | 2021-12-01 | 2022-03-01 | 清华大学 | Multi-target tracking method based on occlusion compensation and used for three-dimensional space information fusion |
| CN114119671B (en) * | 2021-12-01 | 2022-09-09 | 清华大学 | Multi-target tracking method based on occlusion compensation and used for three-dimensional space information fusion |
| CN114373298A (en) * | 2021-12-16 | 2022-04-19 | 苏州思卡信息系统有限公司 | Method for calculating lane number by adopting roadside laser radar |
| CN114373298B (en) * | 2021-12-16 | 2023-02-17 | 苏州思卡信息系统有限公司 | Method for calculating lane number by adopting roadside laser radar |
| CN114494618A (en) * | 2021-12-30 | 2022-05-13 | 广州小鹏自动驾驶科技有限公司 | Map generation method and device, electronic equipment and storage medium |
| CN115937812A (en) * | 2023-01-06 | 2023-04-07 | 河北博士林科技开发有限公司 | Method and system for generating virtual lane line at traffic intersection |
| CN116679293A (en) * | 2023-08-01 | 2023-09-01 | 长沙隼眼软件科技有限公司 | Multi-radar target track splicing method and device based on high-precision map |
| CN116679293B (en) * | 2023-08-01 | 2023-09-29 | 长沙隼眼软件科技有限公司 | Multi-radar target track splicing method and device based on high-precision map |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113537046A (en) | Map lane marking method and system based on vehicle track big data detection | |
| CN111551958B (en) | Mining area unmanned high-precision map manufacturing method | |
| EP4145393B1 (en) | Vehicle localization | |
| EP3343503B1 (en) | High-precision map data processing method and apparatus, storage medium and device | |
| CN109643367B (en) | Methods, systems, and readable media for crowd sourcing and distributing sparse maps and lane measurements for autonomous vehicle navigation | |
| CN111448478B (en) | System and method for correcting high-definition maps based on obstacle detection | |
| CN107235044B (en) | A kind of restoring method realized based on more sensing datas to road traffic scene and driver driving behavior | |
| CN105302151B (en) | A kind of system and method for aircraft docking guiding and plane type recognition | |
| Zhao et al. | On-road vehicle trajectory collection and scene-based lane change analysis: Part i | |
| CN110765894B (en) | Target detection method, device, equipment and computer readable storage medium | |
| US20190310651A1 (en) | Object Detection and Determination of Motion Information Using Curve-Fitting in Autonomous Vehicle Applications | |
| CN113255520B (en) | Vehicle obstacle avoidance method based on binocular vision and deep learning and electronic equipment | |
| CN106584451B (en) | automatic transformer substation composition robot and method based on visual navigation | |
| CN110598637B (en) | Unmanned system and method based on vision and deep learning | |
| CN109583415A (en) | A kind of traffic lights detection and recognition methods merged based on laser radar with video camera | |
| CN111958592A (en) | Image semantic analysis system and method for transformer substation inspection robot | |
| CN103954275A (en) | Lane line detection and GIS map information development-based vision navigation method | |
| CN111006655A (en) | Multi-scene autonomous navigation positioning method for airport inspection robot | |
| CN114862901A (en) | Road-end multi-source sensor fusion target sensing method and system for surface mine | |
| CN110986945B (en) | Local navigation method and system based on semantic altitude map | |
| CN117576652B (en) | Road object identification method and device, storage medium and electronic equipment | |
| CN114808649B (en) | Highway scribing method based on vision system control | |
| CN115564865A (en) | Construction method and system of crowdsourcing high-precision map, electronic equipment and vehicle | |
| CN115562076B (en) | Simulation system, method and storage medium for unmanned mine car | |
| Xia et al. | CyberC3: A prototype cybernetic transportation system for urban applications |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20211022 |
|
| WW01 | Invention patent application withdrawn after publication |