CN110956072B - Driving skill training method based on big data analysis - Google Patents
Driving skill training method based on big data analysis Download PDFInfo
- Publication number
- CN110956072B CN110956072B CN201910698363.XA CN201910698363A CN110956072B CN 110956072 B CN110956072 B CN 110956072B CN 201910698363 A CN201910698363 A CN 201910698363A CN 110956072 B CN110956072 B CN 110956072B
- Authority
- CN
- China
- Prior art keywords
- driving
- data
- training
- driver
- new
- 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
- 238000012549 training Methods 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000007405 data analysis Methods 0.000 title claims abstract description 9
- 230000006399 behavior Effects 0.000 claims abstract description 66
- 206010039203 Road traffic accident Diseases 0.000 claims abstract description 25
- 238000004458 analytical method Methods 0.000 claims abstract description 25
- 238000001514 detection method Methods 0.000 claims description 36
- 230000008569 process Effects 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 19
- 238000003062 neural network model Methods 0.000 claims description 14
- 238000005516 engineering process Methods 0.000 claims description 13
- 238000007781 pre-processing Methods 0.000 claims description 13
- 238000004891 communication Methods 0.000 claims description 12
- 238000002372 labelling Methods 0.000 claims description 9
- 230000002093 peripheral effect Effects 0.000 claims description 6
- 238000007418 data mining Methods 0.000 claims description 5
- 230000010354 integration Effects 0.000 claims description 5
- 238000004140 cleaning Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 4
- 238000011049 filling Methods 0.000 claims description 4
- 238000000611 regression analysis Methods 0.000 claims description 4
- 238000012098 association analyses Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 2
- 238000009412 basement excavation Methods 0.000 claims 1
- 239000000284 extract Substances 0.000 claims 1
- 230000001133 acceleration Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000007619 statistical method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
- G06V20/597—Recognising the driver's state or behaviour, e.g. attention or drowsiness
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
- G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
- G09B9/00—Simulators for teaching or training purposes
- G09B9/02—Simulators for teaching or training purposes for teaching control of vehicles or other craft
- G09B9/04—Simulators for teaching or training purposes for teaching control of vehicles or other craft for teaching control of land vehicles
- G09B9/052—Simulators for teaching or training purposes for teaching control of vehicles or other craft for teaching control of land vehicles characterised by provision for recording or measuring trainee's performance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Multimedia (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Aviation & Aerospace Engineering (AREA)
- Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Educational Technology (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a driving skill training method based on big data analysis, which comprises the following steps of firstly, efficiently identifying dangerous driving behaviors of a driver: classifying and counting driving data of a driving vehicle when traffic accidents and traffic violations are generated, which are provided by traffic management departments of public security parts, and recognizing dangerous driving behaviors to form traffic management driving big data; then, intelligent training of driving skills is carried out on the new driver by using an intelligent robot coach in a driving school, and classification statistics is carried out on driving behaviors of the new driver to form driving training driving big data; finally, the two types of driving big data are analyzed and processed to obtain which driving behaviors of the new driver belong to dangerous driving, and the dangerous driving behaviors of the new driver are corrected and guided by the robot intelligent coach according to the analysis result, so that the training quality of driving schools is improved, and the driving skill of the driver is improved.
Description
Technical Field
The invention relates to a training method, in particular to a driving skill training method capable of effectively avoiding dangerous driving based on big data analysis.
Background
The national security administration and administration at the end of 2017 and the road traffic safety development report jointly issued by the transportation department show that the national total reported road traffic accidents 864.3 are thousands of, and rise by 16.5% as compared with the national total reported road traffic accidents, so that 6.3 thousands of people die, and the number of dead people in the national road traffic accidents is still high in the second place in the world. Although the road traffic environment of China is increasingly improved in recent years, traffic accidents are still high, and the serious road is still left to prevent the high occurrence of the road traffic accidents and reduce the damage of the traffic accidents. According to the data published by the public security department, the traffic accident count of new drivers with driving age within 3 years accounts for more than 45% of the total social accident count, more than 2500 thousands of new drivers are added nationwide in 2015-2018 years, more than 7500 thousands of new drivers are accumulated in the last three years, the average traffic violation is 1.4 times, the annual traffic violation is more than 1 hundred million times, the annual traffic accident count is more than 400 tens of thousands, and the economic loss is more than 1000 hundred million. Based on the basic national conditions of China, more than 98% of driving skills of motor vehicle drivers are learned in driving schools, and the key is to improve the driving skills of the drivers and the safety driving consciousness and the training quality of the driving schools. At present, the phenomenon of the driving training test teaching in China is common, and the driving training method is seriously disjointed with the driving of an actual road. In the next 10 years, new drivers in China can be kept above 2000 ten thousand each year, thousands of new drivers are on the road, so that road traffic safety is under huge pressure, the road driving skill of the new drivers is improved, and the safety driving consciousness of the new drivers is improved.
Disclosure of Invention
The invention aims to: the driving skill training method based on the big data driver behavior analysis is provided, so that the training quality of driving schools is improved, the driving skill of drivers is improved, the establishment of traffic policies and the design of driving examination subjects are improved and perfected, the road traffic safety level is improved, and higher guarantee is brought to national road traffic safety.
The technical scheme is as follows: in order to achieve the above purpose, the invention adopts the following technical scheme: firstly, dangerous driving behaviors of a driver are identified efficiently: the driving data of the driving vehicles when traffic accidents and traffic violations are generated, which are provided by traffic management departments of public security parts, are classified and counted, and the driving behavior risk degree is identified by analyzing the driving track of the vehicles and the traffic conditions at the moment, and dangerous driving behaviors are identified. Then, intelligent training of driving skills is carried out on the new driver by using an intelligent robot coach in a driving school; the robot intelligent coach auxiliary driving system collects driving data of a new driver in the training process through environment sensing devices such as satellite positioning, image recognition, radar detection and vehicle-mounted sensing, guides the new driver to drive correctly, normally and safely in real time, and performs classification statistics on driving behaviors and vehicle driving tracks of the new driver. Finally, the two types of driving big data are statistically analyzed, the driving track of the vehicle is compared, the driving behaviors of the driver are analyzed, and the error driving operation of which new drivers can cause traffic violations and even induce traffic accidents is obtained; and finally, correcting and training dangerous driving behaviors of the new driver through the intelligent robot coach according to the analysis result, so as to avoid the occurrence of the dangerous driving behaviors.
Specifically, the invention discloses a driving skill training method based on big data analysis, which comprises the following steps:
the dangerous driving behavior of the driver is identified with high efficiency: classifying and counting driving data of related driving vehicles provided by traffic management departments of public security parts, and identifying dangerous driving behaviors by analyzing current traffic conditions and driving tracks of the vehicles to form a traffic management driving large database;
intelligent training is carried out on the driving behaviors of the new driver: the intelligent training of the new driver is carried out by using the intelligent robot coach auxiliary driving system, the safe driving of the new driver is guided in real time, and the intelligent robot coach auxiliary driving system collects driving data of the new driver in the training process to form a driving training driving database;
analyzing and processing traffic management and driving training driving big data, and correcting dangerous driving behaviors of new drivers: and comparing the running track of the vehicle, analyzing the driving behaviors of the driver, obtaining which driving behaviors of the new driver belong to dangerous driving, and correcting the dangerous driving behaviors of the new driver through the robot intelligent coach auxiliary driving system according to the analysis result, so as to avoid the dangerous driving behaviors.
Further, in the process of efficiently identifying dangerous driving behaviors of a driver, a region of interest of image information provided by a public security department is intercepted by combining prior knowledge to serve as a characteristic part, a deep learning method is used for carrying out target identification on the characteristic part, and the identification rate of a related vehicle driving track is greatly improved by designing a convolutional neural network.
Further, the target recognition process specifically includes the following steps:
step S1, preprocessing an acquired image group containing a target to be identified, recording the processed image group as an image set img, and dividing the image set img into a training set and a verification set;
s2, arranging region labeling information contained in each image in an image set img as { whether the region is a background, the central coordinates of a boundary frame of the region, the length of the boundary frame of the region, the width of the boundary frame of the region }, and marking as a labeling set label, wherein the range of the boundary frame labeling of the region comprises a target to be identified;
s3, constructing a neural network model as Y (img) =label, putting the image set img and the label set label into the neural network model Y for training, and obtaining parameters of the neural network model Y through training;
and S4, putting the preprocessed recognition scene image new_img into a trained neural network model Y to obtain a target position label new_label to be recognized in the recognition scene image, and cutting the target image to be recognized according to the new_label to realize recognition of the target.
Further, the robot intelligent coach auxiliary driving system includes: the detection unit and the peripheral unit are respectively connected with the control unit, and the control unit comprises a processor and a memory; the detection unit comprises a positioning detection device, a radar detection device and an image detection device; the peripheral unit includes a communication device, a display device, a speaker, a microphone.
Further, the robot intelligent coaching assist driving system further includes a roof apparatus including a bracket 1, and a mechanical lidar 2, a first satellite antenna 3 and a second satellite antenna 4, and a first solid-state lidar 5 and a second solid-state lidar 6 arranged on the bracket 1.
Further, the mechanical lidar 2 is located at the middle position of the bracket 1, the mechanical lidar 2 is located between the first satellite antenna 3 and the second satellite antenna 4, the first satellite antenna 3 and the first solid-state lidar 5 are located at one end of the bracket 1, the second satellite antenna 4 and the second solid-state lidar 6 are located at the other end of the bracket 1, and the solid- state lidars 5 and 6 are located at the outermost sides of the bracket 1 respectively.
Further, the process of analyzing and processing traffic management and driving training big data adopts Hadoop distributed system infrastructure technology to expand and package, the provided data is analyzed through python language and tensorsurface symbol mathematical system, the analysis result is used for defining and creating the optimal parameters of the mining model, the optimal parameters are applied to the whole data set, and the optional modes and the statistical information are extracted.
Further, the process of analyzing and processing traffic management and driving training big data specifically comprises the following steps:
(1) And (3) data acquisition: big data acquisition, database acquisition and file acquisition;
(2) Data preprocessing: data cleaning, data integration, data conversion, data protocol and data integration;
(3) And (3) data processing: aiming at multidimensional data of a driver in traffic management and driving training data, connecting a hive database by using a python language and a tensorsurface symbol mathematical system to acquire data, carrying out data mining by association analysis and regression analysis, analyzing the relation between dangerous driving behaviors in the traffic management data and new driving behaviors of the driving training data, establishing a prediction model according to analysis results, further analyzing and predicting driving data acquired in real time, predicting the situation of traffic accidents possibly happening to the driver, and finally transmitting the prediction results to a driving school platform database by using an sqoop tool;
(4) And (3) data output: the driving school management platform issues a driver behavior prediction result and a correction scheme through a WEB server and/or a mobile terminal APP.
Further, the database acquisition comprises the step of importing basic data of a driving training data platform into a hive database of a big data platform through an sqoop tool; the file acquisition comprises the steps of inquiring and writing traffic management data into a local log file through a traffic management department interface at regular time, and writing driving training data acquired through a Beidou positioning device, a vehicle-mounted sensor and a computer vision acquisition device into a background server log file through a driving training platform background interface.
Further, the data preprocessing step further comprises preprocessing data collected by the HDFS distributed file system by writing a MapReduce program, processing and converting missing data, noise data and inconsistent data according to a data protocol, generating a new data file, storing the new data file into the HDFS system, filling the missing data with global constants, attribute means and possible values, and removing the noise data by using a binning, clustering, manual checking by a computer and regression method.
The beneficial effects are that:
the invention provides a driving skill training method based on data driver behavior analysis, which improves the training quality of drivers and brings higher guarantee to road traffic safety; is beneficial to improving and perfecting the formulation of traffic policy and the design of driving examination subjects, and improves the road traffic safety level.
The technology solves the road driving safety problem that the existing driving training method is seriously separated from the actual road driving environment, reduces the national traffic accident rate from the driving safety source, and practically ensures the road traffic safety of China.
Drawings
FIG. 1 is a block diagram of a robot intelligent coaching assist drive system.
Fig. 2 is a diagram of other detecting units.
Fig. 3 is a diagram of a fleet apparatus of a robot intelligent trainer assisted driving system.
Fig. 4 is a flow chart of the big data processing of the present invention.
Detailed Description
The invention will be further illustrated with reference to examples.
Firstly, dangerous driving behaviors of a driver are identified efficiently: classifying and counting driving data of a running vehicle when traffic accidents and traffic violations occur, which are provided by traffic management departments of public security departments, and mainly counting the driving data of the traffic accidents caused by illegal lane changing, rear-end collision, red light running and the like; and carrying out behavior risk recognition on the driving behavior by analyzing the running track of the vehicle and the traffic condition at the time, and recognizing dangerous driving behavior. When classifying and counting driving data provided by public security departments, how to establish a corresponding analysis model according to different driving behaviors and how to realize efficient identification of a target object under different road environments are always important in technical research. Firstly, intercepting an interested region of image information provided by a public security department by combining prior knowledge as a characteristic part, carrying out target identification on the characteristic part by using a deep learning method, and greatly improving the identification rate of the related vehicle driving track by designing a convolutional neural network. Taking the establishment of a rear-end collision track analysis model as an example, establishing a lane-changing track analysis model on the basis of a fuzzy logic model, a neural network model and a motion wave hybrid model, and establishing the rear-end collision track analysis model by utilizing a single-factor variance analysis method.
The specific recognition method in the process of recognizing the image target comprises the following steps:
step S1, preprocessing an acquired image group containing a target to be identified, recording the processed image group as an image set img, and dividing the image set img into a training set and a verification set;
s2, arranging region labeling information contained in each image in an image set img as { whether the region is a background, the central coordinates of a boundary frame of the region, the length of the boundary frame of the region, the width of the boundary frame of the region }, and marking as a labeling set label, wherein the range of the boundary frame labeling of the region comprises a target to be identified;
s3, constructing a neural network model as Y (img) =label, putting the image set img and the label set label into the neural network model Y for training, and obtaining parameters of the neural network model Y through training;
and S4, putting the preprocessed recognition scene image new_img into a trained neural network model Y to obtain a target position label new_label to be recognized in the recognition scene image, and cutting the target image to be recognized according to the new_label to realize recognition of the target.
Further, in step S3, the neural network includes convolution, activation, pooling operations.
Further, in step S3, the neural network model uses the loss function for the class of training recognition target positions in Y as:
wherein m is the number of samples, i is 1 to m; k is the number of categories, l is 1 to k; e is a natural constant.
Further toIn step S3, the loss function for regressing the recognition target position bounding box is loss= |y (img) -label|| 2 。
Further, the identification method further comprises the following steps: and (3) repeating the steps S1-S4 on the image which is identified in the step S4 and contains the secondary target to be identified, so as to identify the secondary target.
Further, in order to improve the high-precision recognition of the smaller recognition target, the steps S1-S4 may be repeated multiple times until the recognition precision meets the requirement.
The driving training is carried out on the new driver by using a robot intelligent coach in a driving school; the intelligent robot coach auxiliary driving system collects driving data of a new driver in the training process through devices such as satellite positioning, radar detection, image recognition and vehicle-mounted sensing, and uneven driving skills of students can be caused due to the fact that different coaches exist, and the traditional driving school mainly improves the training quality of the driver through gold coach or corrects the driving operation of the driver through a training video playback mode. However, the training amount of the gold card coach is limited, the video playback has guiding lag, and the driving skills of students cannot be generally improved. The intelligent robot training system integrates environment sensing technologies such as satellite positioning, radar detection, image recognition, vehicle-mounted sensing and the like, constructs a high-precision lane model on the basis of a vehicle contour model and a driving road model, analyzes the position relationship between the vehicle model and the road model in real time, reminds a driver to make correct operation at key driving moments, and uses the most universal and standard driving operation to train and teach the driving skills of a large number of students on the real-time road and on the site, so that the driving skills and the safety consciousness of new drivers are greatly improved, and the training effect of a large number of driving schools is also improved. In the training process, the robot intelligent coaching system classifies and counts the driving characteristic behaviors of the new driver to form a driving-school training driving track database, and uploads the driving track database to the cloud server.
As shown in fig. 1, the robot intelligent coaching assist driving system of the present invention includes: the detection unit and the peripheral unit are respectively connected with the control unit, and the control unit comprises a processor and a memory; the detection unit comprises a positioning detection device, a radar detection device and an image detection device; the peripheral unit comprises a communication device, a display device, a loudspeaker and a microphone; the driving assistance system further comprises a power supply unit for supplying power to the device.
As shown in fig. 2, the detection unit further includes other vehicle-mounted detection devices including one or a combination of steering, gear, brake, throttle, lights, clutch, horn, hand brake, wiper, door, ignition, and belt detection devices. The detection unit further comprises a fingerprint detector and/or a card reader for detecting the identity of the driver.
The positioning detection device comprises a satellite navigation positioning system and/or an inertial navigation positioning system. The satellite navigation positioning system (GNSS) adopts an RTK differential satellite positioning system, and can be one or a combination of a Global Positioning System (GPS), a Geronus satellite navigation system (GLONASS), a Galileo Satellite Navigation System (GSNS) and a Beidou satellite navigation system (BDS). The satellite positioning device comprises two satellite signal receiving antennas which are arranged on the top of the vehicle, and the distance between the satellite signal receiving antennas is larger than 1 meter. The inertial navigation positioning system (IMU) comprises three single-axis accelerometers and three single-axis gyroscopes, wherein the accelerometers detect acceleration signals of the object in the carrier coordinate system in an independent triaxial mode, the gyroscopes detect angular velocity signals of the carrier relative to the navigation coordinate system, angular velocity and acceleration of the object in a three-dimensional space are measured, and the gesture of the object is calculated according to the angular velocity and the acceleration signals, so that the defects of satellite navigation positioning can be overcome.
The radar detection device includes one of an ultrasonic radar detection device, a millimeter wave radar detection device, a laser radar detection device, or a combination thereof. Wherein, ultrasonic radar is installed in the vehicle outside periphery for detecting object information in the vehicle periphery short distance. The millimeter wave radar is arranged at the front and/or rear of the vehicle and is used for detecting objects within a certain distance range in front and behind the vehicle. The laser radar detection device further comprises a mechanical laser radar detection device and/or a solid-state laser radar detection device. The mechanical laser radar is used for detecting object information in a 360-degree range around the vehicle, and 4-line, 8-line, 16-line, 32-line and 64-line laser radars are selected, the laser radars are combined with GNSS and IMU to scan the 360-degree laser around the vehicle, and the obtained data contains space three-dimensional information and laser intensity information. Solid-state lidar is used to detect object information within a range of a vehicle.
The communication equipment comprises local area network communication equipment and/or mobile network communication equipment, wherein the local area network communication adopts radio station communication or WIFI communication, and the mobile network communication adopts the technology of more than 3G to support full network communication (mobile, communication, telecommunication and the like).
The image detection device comprises an image acquisition device and an image recognition device, wherein the image acquisition device comprises a first camera for acquiring road image information in front of a vehicle, the camera can be arranged at the position near a vehicle rearview mirror or above a center console, and the image recognition device corresponding to the first camera carries out classification recognition on pedestrians, vehicles, lane lines, traffic lights and traffic signs on the road and sends recognition results to the control unit. The image acquisition device further comprises a second camera for acquiring image information of the driving position in the vehicle, the camera can be arranged at a position near the rearview mirror in the vehicle or at a position above the center console of the main driving position, and the image recognition device corresponding to the second camera recognizes whether a driver observes left and right in the environments of starting, passing through an intersection and the like. The image detection device includes a third camera that collects image information of the road behind the vehicle, and the camera may be provided at a position near the bumper behind the vehicle. The image detection device comprises a fourth camera for collecting road image information on the left side and the right side of the vehicle, and the camera can be arranged at the lower position of the outside rearview mirror. The image detection device further comprises a fifth camera for acquiring image information of the copilot, and the camera can be arranged at a position near the rearview mirror in the vehicle. The image detection device further comprises a sixth camera for collecting hand operations of the driver and is responsible for identifying gear engaging operations and steering wheel operations of the driver, and the camera can be arranged at a position near the rearview mirror in the vehicle or at the front-row roof position.
The robot smart trainer assisted driving system further includes a roof device, as shown in fig. 3, which includes a stand 1, and a mechanical lidar 2, a first satellite antenna 3 and a second satellite antenna 4, and a first solid-state lidar 5 and a second solid-state lidar 6, which are arranged on the stand 1. The mechanical laser radar 2 is positioned at the middle position of the bracket 1, and collects environmental information data around the outside of the vehicle and is arranged between the first satellite antenna 3 and the second satellite antenna 4. The first satellite antenna 3 and the first solid-state lidar 5 are located at one end of the bracket 1, and the second satellite antenna 4 and the second solid-state lidar 6 are located at the other end of the bracket 1, wherein the solid- state lidars 5, 6 are located at the outermost sides of the bracket 1, respectively. Wherein, the first solid-state laser radar 5 and the second solid-state laser radar 6 collect road environment information data of the side and/or the front of the vehicle, and the first satellite antenna 3 and the second satellite antenna 4 are used for receiving satellite positioning signals.
Through the two types of driving big data provided by public security departments and wide driving schools, the error driving operation of which new drivers can cause traffic violations and even induce traffic accidents is obtained through analysis; and correcting related dangerous driving behaviors of the new driver through the robot intelligent coach according to the analysis result, improving the road, site driving skills and safe driving consciousness of the new driver, and avoiding the occurrence of dangerous driving behaviors. Traditional driver driving behavior analysis mainly analyzes driving behaviors causing traffic accidents or traffic violations from a macroscopic perspective, and does not perform statistical analysis on specific driving operations of microscopic drivers. The method comprises the steps of clustering and analyzing a real road driving track database and a driving-school training driving track database, identifying the risk degree of similar driving tracks, establishing a Logit improved model of the basic driving operation characteristics of a new driver on the influence of responsible traffic accidents or traffic violations, analyzing the correlation between the basic driving operation characteristics of the new driver and dangerous driving behaviors and serious consequences thereof, determining the basic driving operation with the dangerous driving behaviors, and facilitating the intelligent robot coaching system to correct the dangerous operation of the new driver in time in the driving school.
The big data processing technology can be expanded and packaged based on the Hadoop technology, and the process of the relevant big data technology is derived by utilizing the advantages of Hadoop open source and relevant characteristics (good at processing unstructured, semi-structured data, complex ETL flow, complex data mining and calculation model and the like) aiming at data and scenes (storage, calculation and the like of unstructured data) which are difficult to process by a traditional relational database; the data provided by python and tensorflow analysis is then looked up for analysis methods such as correlation analysis, regression analysis, etc., and the analysis results are used to define the best parameters for creating the mining model and these parameters are applied to the entire dataset to extract the feasible patterns and detailed statistical information. The big data processing flow of the invention is shown in figure 4:
(1) Data acquisition
Big data acquisition, namely, acquisition of structured and unstructured mass data of various sources (such as traffic management data, driving training data and the like).
And (3) collecting a database: and importing the basic data of the driving training data platform into a hive database of a large data platform through an sqoop tool.
And (3) file collection: the traffic management data is periodically queried and written into a local log file through a traffic management department interface; the vehicle-mounted data are used for acquiring data of a driver in the driving training process from different angles through Beidou positioning, vehicle-mounted sensors and computer vision, and the data are written into a background server log file through a background interface of the driving training platform. The server installation flime system monitoring logs of each log storage are transmitted to the big data server flime system through the avro system, and the big data server flime system stores data into the HDFS system and the message queues written into the kafka system respectively.
(2) Data preprocessing
The data preprocessing refers to a series of operations such as cleaning, filling, smoothing, merging, normalizing, consistency checking and the like, which are performed on the collected original data before data analysis, and aims to improve the data quality and lay a foundation for later analysis work. The data preprocessing mainly comprises four parts: data cleaning, data integration, data conversion and data protocol.
Preprocessing data collected by the HDFS by writing a MapReduce program, processing and converting missing data (lack of interesting attributes), noise data (data with errors or deviation from expected values) and inconsistent data according to a data protocol, and generating a new data file to store in the HDFS. The missing data processing method comprises the following steps: filling with global constants, attribute means, possible values; or directly ignore the data; the noise data processing method comprises the following steps: noise treatment is removed by the methods of box division (raw data are grouped and each group of data is processed smoothly), clustering, manual inspection by a computer, regression and the like.
And (3) integrating data, creating a scheduling task through an oozie scheduling engine, periodically creating a partition to the hive data table according to the date, pointing to an HDFS file, and calling sqoop to import driving training basic data increment into the hive table.
(3) Data processing
Through the research of technologies such as distributed storage and retrieval of data, data mining-oriented distributed computing framework selection and fusion, a driver behavior analysis prediction model based on big data is built on the basis of system design and data mining. Aiming at multidimensional data of drivers in traffic management and driving training big data, a python language and a tensorsurface symbol mathematical system are used for connecting a hive database to acquire data, data mining is carried out through technologies such as association analysis and regression analysis, the relation between the two is found out, thus valuable association rules are obtained, the relation between traffic accidents and driving training behaviors is analyzed, a prediction model is built according to the results, further, the driving data acquired in real time by a kafka system are analyzed and predicted through the atom on-line processing, the situation of the traffic accidents possibly happened by the drivers in different driving modes is predicted, and the prediction results are transmitted to a driving school management platform database through sqoop.
(4) Data output
The driving school management platform issues a driver behavior prediction result and a correction scheme through a WEB server and/or a mobile terminal APP.
The driving skill training method based on big data analysis firstly carries out high-efficiency identification on dangerous driving behaviors of a driver, carries out classification statistics on driving data of a driving vehicle when traffic accidents and traffic violations are generated, provided by traffic management departments of public security parts, carries out driving behavior risk identification on the driving behaviors by analyzing the driving track of the vehicle and the traffic conditions at the time, and identifies the dangerous driving behaviors. Then, intelligent training of driving skills is carried out on the new driver by using an intelligent robot coach in a driving school; the robot intelligent coach auxiliary driving system collects driving data of a new driver in the training process through environment sensing devices such as satellite positioning, image recognition, radar detection and vehicle-mounted sensing, guides the new driver to drive correctly, normally and safely in real time, and performs classification statistics on driving behaviors and vehicle driving tracks of the new driver. Finally, processing and analyzing the two types of driving big data, comparing the driving track of the vehicle and analyzing the driving behavior of the driver to obtain which new drivers can cause traffic violations and even induce traffic accidents; and finally, correcting and training dangerous driving behaviors of the new driver through the intelligent robot coach according to the analysis result, so as to avoid the occurrence of the dangerous driving behaviors.
The driving skill training method based on big data analysis mainly adopts the image recognition technology based on deep learning, the auxiliary driving technology based on the robot intelligent coach and the big data processing technology to carry out statistical analysis on driving behaviors causing traffic accidents and traffic violations and driving behaviors of new drivers in the training process, and carries out correction training on dangerous driving behaviors of the drivers, thereby improving the driving skill and safety driving consciousness of the drivers.
Through statistical analysis of training data of students of more than 1500 driving schools nationwide on a vehicle learning platform, the driving behaviors of a new driver in the training process of a real vehicle and a simulator are subjected to risk assessment, dangerous driving behaviors are corrected and trained, the driving skill of the new driver is practically improved, the national traffic accident rate is reduced from the driving safety source, and the road traffic safety of China is improved.
Although the embodiments of the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to the above-described specific embodiments and application fields, and the above-described specific embodiments are merely illustrative, and not restrictive. Those skilled in the art, having the benefit of this disclosure, may effect numerous forms of the invention without departing from the scope of the invention as claimed.
Claims (8)
1. The driving skill training method based on big data analysis is characterized by comprising the following steps:
the dangerous driving behavior of the driver is identified with high efficiency: classifying and counting driving data of related driving vehicles provided by traffic management departments of public security parts, and identifying dangerous driving behaviors by analyzing current traffic conditions and driving tracks of the vehicles to form a traffic management driving large database;
intelligent training is carried out on the driving behaviors of the new driver: the intelligent training of the new driver is carried out by using the intelligent robot coach auxiliary driving system, the safe driving of the new driver is guided in real time, and the intelligent robot coach auxiliary driving system collects driving data of the new driver in the training process to form a driving training driving database;
analyzing and processing traffic management and driving training driving big data, and correcting dangerous driving behaviors of new drivers: comparing the running track of the vehicle, analyzing the driving behaviors of the driver to obtain which driving behaviors of the new driver belong to dangerous driving, and correcting the dangerous driving behaviors of the new driver through the robot intelligent coach auxiliary driving system according to the analysis result to avoid the dangerous driving behaviors;
the analysis processing management and driving training big data process adopts Hadoop distributed system infrastructure technology to expand and package, analyzes provided data through python language and tensorsurface symbol mathematical system, defines and creates optimal parameters of an excavation model by using analysis results, applies the optimal parameters to the whole data set, and extracts optional modes and statistical information;
the process for analyzing and processing traffic management and driving training big data specifically comprises the following steps:
(1) And (3) data acquisition: big data acquisition, database acquisition and file acquisition;
(2) Data preprocessing: data cleaning, data integration, data conversion, data protocol and data integration;
(3) And (3) data processing: aiming at multidimensional data of a driver in traffic management and driving training data, connecting a hive database by using a python language and a tensorsurface symbol mathematical system to acquire data, carrying out data mining by association analysis and regression analysis, analyzing the relation between dangerous driving behaviors in the traffic management data and new driving behaviors of the driving training data, establishing a prediction model according to analysis results, further analyzing and predicting driving data acquired in real time, predicting the situation of traffic accidents possibly happening to the driver, and finally transmitting the prediction results to a driving school platform database by using an sqoop tool;
(4) And (3) data output: the driving school management platform issues a driver behavior prediction result and a correction scheme through a WEB server and/or a mobile terminal APP.
2. The driving skill training method as set forth in claim 1, wherein in the process of efficiently recognizing dangerous driving behaviors of a driver, a region of interest of image information provided by a police department is intercepted as a characteristic part, a deep learning method is used for carrying out target recognition on the characteristic part, and recognition rate of a related vehicle driving track is greatly improved by designing a convolutional neural network.
3. The driving skill training method according to claim 2, wherein the target recognition process specifically includes the steps of:
step S1, preprocessing an acquired image group containing a target to be identified, recording the processed image group as an image set img, and dividing the image set img into a training set and a verification set;
s2, arranging region labeling information contained in each image in an image set img as { whether the region is a background, the central coordinates of a boundary frame of the region, the length of the boundary frame of the region, the width of the boundary frame of the region }, and marking as a labeling set label, wherein the range of the boundary frame labeling of the region comprises a target to be identified;
s3, constructing a neural network model as Y (img) =label, putting the image set img and the label set label into the neural network model Y for training, and obtaining parameters of the neural network model Y through training;
and S4, putting the preprocessed recognition scene image new_img into a trained neural network model Y to obtain a target position label new_label to be recognized in the recognition scene image, and cutting the target image to be recognized according to the new_label to realize recognition of the target.
4. The driving skill training method according to claim 1, wherein the robot intelligent trainer assisted driving system comprises: the detection unit and the peripheral unit are respectively connected with the control unit, and the control unit comprises a processor and a memory; the detection unit comprises a positioning detection device, a radar detection device and an image detection device; the peripheral unit includes a communication device, a display device, a speaker, a microphone.
5. The driving skill training method according to claim 4, wherein the robot intelligent trainer assisted driving system further comprises a roof device, the roof device comprises a bracket 1 and a mechanical laser radar (2), a first satellite antenna (3), a second satellite antenna (4), a first solid-state laser radar (5) and a second solid-state laser radar (6) which are arranged on the bracket (1).
6. The driving skill training method according to claim 5, wherein the mechanical lidar (2) is located at a position just in the middle of the support (1), the mechanical lidar (2) is located between the first satellite antenna (3) and the second satellite antenna (4), the first satellite antenna (3) and the first solid-state lidar (5) are located at one end of the support (1), the second satellite antenna (4) and the second solid-state lidar (6) are located at the other end of the support (1), and the solid-state lidars (5), (6) are located at the outermost sides of the support (1), respectively.
7. The driving skill training method of claim 1, wherein the database collection comprises importing basic data of a driving skill data platform into a big data platform hive database through an sqoop tool; the file acquisition comprises the steps of inquiring and writing traffic management data into a local log file through a traffic management department interface at regular time, and writing driving training data acquired through a Beidou positioning device, a vehicle-mounted sensor and a computer vision acquisition device into a background server log file through a driving training platform background interface.
8. The driving skill training method according to claim 1, wherein the data preprocessing step further comprises preprocessing data collected by the HDFS distributed file system by writing a MapReduce program, processing and converting missing data, noise data and inconsistent data according to a data protocol, generating a new data file, storing the new data file in the HDFS system, filling the missing data with global constants, attribute means and possible values, and removing the noise data by a binning, clustering, manual inspection by a computer and regression method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910698363.XA CN110956072B (en) | 2019-07-31 | 2019-07-31 | Driving skill training method based on big data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910698363.XA CN110956072B (en) | 2019-07-31 | 2019-07-31 | Driving skill training method based on big data analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110956072A CN110956072A (en) | 2020-04-03 |
| CN110956072B true CN110956072B (en) | 2023-06-02 |
Family
ID=69976209
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910698363.XA Active CN110956072B (en) | 2019-07-31 | 2019-07-31 | Driving skill training method based on big data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110956072B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112348095A (en) * | 2020-11-10 | 2021-02-09 | 易显智能科技有限责任公司 | Method and related device for evaluating safe driving consciousness of driver |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103971558A (en) * | 2014-05-16 | 2014-08-06 | 无锡海鸿信息技术有限公司 | Intelligent robot system for motor vehicle driver training |
| CN108545019A (en) * | 2018-04-08 | 2018-09-18 | 多伦科技股份有限公司 | A kind of safety driving assist system and method based on image recognition technology |
-
2019
- 2019-07-31 CN CN201910698363.XA patent/CN110956072B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103971558A (en) * | 2014-05-16 | 2014-08-06 | 无锡海鸿信息技术有限公司 | Intelligent robot system for motor vehicle driver training |
| CN108545019A (en) * | 2018-04-08 | 2018-09-18 | 多伦科技股份有限公司 | A kind of safety driving assist system and method based on image recognition technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110956072A (en) | 2020-04-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107782564B (en) | Automatic driving vehicle evaluation system and method | |
| CN111179585B (en) | Site testing method and device for automatic driving vehicle | |
| CN112700470B (en) | Target detection and track extraction method based on traffic video stream | |
| WO2020042984A1 (en) | Vehicle behavior detection method and apparatus | |
| EP3104284A1 (en) | Automatic labeling and learning of driver yield intention | |
| EP3687863A1 (en) | Multiple exposure event determination | |
| CN207624060U (en) | A kind of automated driving system scene floor data acquisition system | |
| CN106680212B (en) | A kind of rectilinear telemetering motor vehicle tail system based on road reflection | |
| CN102881162A (en) | Data processing and fusion method for large-scale traffic information | |
| CN113155173B (en) | Perception performance evaluation method and device, electronic device and storage medium | |
| CN105448105A (en) | Patrol police vehicle-based monitoring system | |
| CN106340205A (en) | Traffic monitoring method and traffic monitoring apparatus | |
| CN106412508A (en) | Intelligent monitoring method and system of illegal line press of vehicles | |
| CN109094574B (en) | Unmanned vehicle driving state measurement control system based on rack | |
| CN113869275A (en) | Vehicle object detection system that throws based on remove edge calculation | |
| CN115291515A (en) | Automatic driving simulation test system and evaluation method based on digital twinning | |
| CN110956072B (en) | Driving skill training method based on big data analysis | |
| CN113076999B (en) | Information data acquisition method based on artificial intelligence | |
| CN110379173A (en) | A kind of Manpower Transportation unlawful practice recognition methods based on shared bicycle multi-modal data | |
| CN105303837A (en) | Method and system for detecting following behavior characteristic parameter of driver | |
| CN112633163B (en) | Detection method for realizing illegal operation vehicle detection based on machine learning algorithm | |
| US20220172606A1 (en) | Systems and Methods for Extracting Data From Autonomous Vehicles | |
| CN114814825B (en) | Vehicle track sensing and state extraction method based on radar and video fusion | |
| Roomi et al. | Deep Learnt Anti Collision System for Highway Safety | |
| Zhang et al. | Virtual validation method for automated driving vehicles based on traffic accident |
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 |