CN110458214A - Driver replaces recognition methods and device - Google Patents
Driver replaces recognition methods and device Download PDFInfo
- Publication number
- CN110458214A CN110458214A CN201910698186.5A CN201910698186A CN110458214A CN 110458214 A CN110458214 A CN 110458214A CN 201910698186 A CN201910698186 A CN 201910698186A CN 110458214 A CN110458214 A CN 110458214A
- Authority
- CN
- China
- Prior art keywords
- data
- driver
- model
- driving
- sample
- 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.)
- Granted
Links
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C5/00—Registering or indicating the working of vehicles
- G07C5/08—Registering or indicating performance data other than driving, working, idle, or waiting time, with or without registering driving, working, idle or waiting time
- G07C5/0808—Diagnosing performance data
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C5/00—Registering or indicating the working of vehicles
- G07C5/08—Registering or indicating performance data other than driving, working, idle, or waiting time, with or without registering driving, working, idle or waiting time
- G07C5/0841—Registering performance data
-
- 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
Abstract
The present invention relates to a kind of drivers to replace recognition methods and device, the method includes the steps: driving data obtaining step acquires the history running data of vehicle by OBD equipment；Feature construction step constructs data set and driving habit feature；The study of model and training step, according to the data set and driving habit latent structure and training machine learning model；The update and deploying step of model.The present invention can accurately analyze the driving behavior habit of current driver's, see whether match with history driving habit, to identify whether to replace driver in conjunction with the external environment of history the traveling record and current driving of vehicle.
Description
Technical field
The present invention relates to car networking and field of artificial intelligence, in particular to a kind of driver replaces recognition methods and dress
It sets.
Background technique
With the development of the social economy, China's vehicles number is continuously increased, the increase of vehicle facilitates the trip of people,
But traffic safety problem is also brought simultaneously.Therefore, for the monitoring of motor-driven vehicle going risk, more stringent requirements are proposed.Its
In, it is a kind of risk relatively conventional and with higher monitoring difficulty that driver, which replaces risk,.
Currently, driver, which replaces identification technology, is broadly divided into two kinds, one is directly onboard install camera, user
Face identification technology identifies driver；Another kind be onboard install OBD (On Board Diagnostics, it is vehicle-mounted to examine automatically
Disconnected system) equipment, driver's running data is acquired, server is passed to, the regulation engine and statistical model established dependent on expert
It is analyzed.
First way is related to the privacy of user, in addition to for there is the client of Fleet size such as motorcade logistic etc. that can obtain
It must permit, be beyond affordability for normal client.The second way uses artificial constructed driving habit feature, this to do
Method compares the experience for relying on expert, heavy workload, and feature database update is comparatively laborious, and accuracy of identification cannot ensure.
Summary of the invention
Based on this, it is necessary to provide a kind of driver and replace recognition methods and device, can be travelled in conjunction with the history of vehicle
The external environment of record and current driving accurately analyzes the driving behavior habit of current driver's, sees whether drive with history
It sails habit to match, to identify whether to replace driver.
For achieving the above object, the present invention uses following technical scheme.
The present invention provides a kind of driver's replacement recognition methods, comprising the following steps:
Driving data obtaining step acquires the history running data of vehicle by OBD equipment；
Feature construction step constructs data set and driving habit feature；
The study of model and training step learn mould according to the data set and driving habit latent structure and training machine
Type；
The update and deploying step of model.
Above-mentioned driver replaces in recognition methods, and the feature construction step further comprises:
Data prediction is decoded the collected history running data of institute in driving data obtaining step, cleans, divides
Class and storage, with excluding outlier and assigning null data；
Data set building constructs sample and data set using pseudo label method；With
Driving habit feature construction, the driving habit based on stroke characteristic, three anxious features and GPS data building driver are special
Sign, the stroke characteristic include stroke time started, stroke end time, drive duration, mileage number, average speed and hypervelocity time
Number；Described three anxious features include anxious acceleration, anxious deceleration and zig zag number and the time point of each run.
Above-mentioned driver replaces in recognition methods, described specifically to be wrapped using the step of pseudo label method building sample and data set
It includes:
Assuming that current vehicle driver does not replace, it, will when constructing sample using the history running data of current vehicle
It is set to 0 as negative sample, and by label；When extracting history running data building sample from other vehicles at random, made
For positive sample, and label is set to 1；
After all negative samples and positive sample are collected, the data set is formed.
The step of above-mentioned driver replaces in recognition methods, the driving habit feature construction specifically includes:
The elongated sequence data in the history running data is converted into fixed length sequence by the way of interception and filling
GPS data in each run is divided into three parts using block sampling method, intercepts K respectively from each part by data
A GPS point, the GPS sequence data that composition length is 3K；
If journey time is shorter or K value is larger causes data to there is intersection, using the mean value for having value to cross part
The null value divided is filled.
Above-mentioned driver replaces in recognition methods, and the study of the model and training step specifically include:
Collection and driving habit feature construction machine learning model based on the data, and model is carried out in the following ways
Study and training:
First the data set and driving habit feature are inputted in a GBDT model, take out the tree that the GBDT model generates
Index；
The mode for reusing one-hot is input in a LR model, exports recognition result by the LR model.
Above-mentioned driver replaces in recognition methods, and the update of the model and deploying step specifically include:
Docker mirror image is constructed on line, the model and prediction code are copied in the docker, when receiving
The history running data of the vehicle is inputted into the docker when history running data of vehicle, is handled by the docker,
Output in real time differentiates result and saves.
The present invention also provides a kind of drivers to replace identification device, comprising:
Driving data obtains module, for acquiring the history running data of vehicle by OBD equipment；
Feature construction module, for constructing data set and driving habit feature；
The study of model and training module, for according to the data set and driving habit latent structure and training machine
Practise model；
The update and deployment module of model, for the machine learning model to be updated and disposed.
Above-mentioned driver replaces in identification device, and the feature construction module further comprises:
Data pre-processing unit, for being solved to the collected history running data of institute in driving data obtaining step
Code, cleaning, classification and storage, with excluding outlier and assigning null data；
Data set construction unit, for using pseudo label method building sample and data set；With
Driving habit feature construction unit, for driving based on stroke characteristic, three anxious features and GPS data building driver
Habit feature is sailed, the stroke characteristic includes the stroke time started, the stroke end time, drives duration, mileage number, average speed
With hypervelocity number；Described three anxious features include anxious acceleration, anxious deceleration and zig zag number and the time point of each run.
Above-mentioned driver replaces in identification device, and the data set construction unit is specifically used for:
Assuming that current vehicle driver does not replace, it, will when constructing sample using the history running data of current vehicle
It is set to 0 as negative sample, and by label；When extracting history running data building sample from other vehicles at random, made
For positive sample, and label is set to 1；
After all negative samples and positive sample are collected, the data set is formed.
Above-mentioned driver replaces in identification device, and driving habit feature construction unit is specifically used for,
The elongated sequence data in the history running data is converted into fixed length sequence by the way of interception and filling
GPS data in each run is divided into three parts using block sampling method, intercepts K respectively from each part by data
A GPS point, the GPS sequence data that composition length is 3K；
If journey time is shorter or K value is larger causes data to there is intersection, using the mean value for having value to cross part
The null value divided is filled.
Recognition methods is replaced compared to traditional driver, the present invention is by obtaining driving data, building data set and driving
Habit feature is sailed, and construction machine learning model not only can protect the privacy of user without installing camera, additionally it is possible to
In conjunction with the external environment of vehicle history traveling record and current driving, the driving behavior for accurately analyzing current driver's is practised
It is used, see whether match with history driving habit, to identify whether to replace driver.And melted by using GBDT+LR model
The mode of conjunction effectively improves the stability and precision of model identification.
Detailed description of the invention
Fig. 1 is the flow diagram that driver replaces recognition methods in the present embodiment；
Fig. 2 is the structural schematic diagram that driver replaces identification device in the present embodiment.
Specific embodiment
With reference to the accompanying drawing and specific embodiment is described further.
As shown in Figure 1, the present embodiment provides a kind of drivers to replace recognition methods, comprising the following steps:
S1: driving data obtaining step acquires the history running data of vehicle by OBD equipment；
S2: feature construction step constructs data set and driving habit feature；
S3: the study of model and training step, according to the data set and driving habit latent structure and training machine
Practise model；
S4: the update and deploying step of model.
When acquiring vehicle data, most simple effective method is to be connect by OBD interface with automobile, and OBD equipment is
Presently the most common OBD diagnosis and data read-write equipment.
Above-mentioned OBD equipment can collect a large amount of related data, such as travel speed of vehicle driving, and running time travels road
Journey, three anxious data (zig zag, anxious deceleration are anxious to accelerate), travels latitude and longitude information etc. a little, is referred to as history traveling herein
Data.
The present embodiment is based on above-mentioned history running data and carries out feature construction, and the feature construction step S1 is further wrapped
It includes:
S21: data prediction is decoded, clearly the collected history running data of institute in driving data obtaining step
It washes, classify and stores, with excluding outlier and assigning null data；
S22: data set building constructs sample and data set using pseudo label method；With
S23: driving habit feature construction, the driving based on stroke characteristic, three anxious features and GPS data building driver are practised
Used feature, when the stroke characteristic includes stroke time started, stroke end time, starting point coordinate, terminal point coordinate, driving
Length, mileage number, average speed, stroke maximum speed, stroke minimum speed, stroke stop interval time, maximum hypervelocity speed and
Exceed the speed limit number, etc.；Described three anxious features include anxious acceleration, anxious deceleration and zig zag number and the time point of each run.
In step S21, due to will receive the influence of many extraneous factors, the collected history of OBD equipment in driving conditions
Running data quality is irregular, such as when crossing-river tunnel, GPS signal is weaker, at this time the GPS number of collected automobile
It is inaccurate according to (such as coordinate and speed), these data can not be obtained sometimes or even at all, this has also resulted in history traveling
Some exceptional values and null value in data.And the data type obtained by OBD interface cannot be used directly, therefore, be needed
Above-mentioned data are pre-processed, including decoding, cleaning, classification and storage, with excluding outlier and assigning null data,
And it is converted into required data format, it saves backup.
After completing above-mentioned steps, the building of data set can be carried out.The advantages of the present embodiment first is that single can be realized
The driver of stroke replaces identification, therefore, in step S22, defines a sample first, defines vehicle starting to from single
Stop working as a stroke, data based on each sample are both from a complete stroke.
When constructing to sample, the present embodiment uses pseudo label method, that is, assuming that current vehicle driver is without more
It changes, when constructing sample using the history running data of current vehicle, is set to 0 as negative sample, and by label；Random
When extracting history running data building sample from other vehicles, 1 is set to as positive sample, and by label.
Then, all negative samples and positive sample are collected, forms the data set.It is easily understood that later use mould
When type verifies data set, when the sample label of a certain run-length data be 0 when, it is known that the driver of the vehicle there is no
Change, when the sample label of a certain run-length data is 1, it is known that the driver of the vehicle is changed.
In step S23, when considering driving habit feature, mainly to stroke time started, stroke end time, starting point
Coordinate, drives duration, mileage number, average speed, stroke maximum speed, stroke minimum speed, the secondary stroke distances at terminal point coordinate
Stroke residence time, maximum hypervelocity speed, hypervelocity number, the three anxious dimensions such as feature and GPS data are analyzed next time,
Wherein, carrying out analysis in the data to above-mentioned each dimension is mainly to be handled GPS data and analyzed, comprising:
The elongated sequence data in the history running data is converted into fixed length sequence by the way of interception and filling
GPS data in each run is divided into three parts using block sampling method, intercepts K respectively from each part by data
A GPS point, the GPS sequence data that composition length is 3K；
Specifically, interception is divided into three parts, and stroke early, middle, late stage intercepts the number of strokes of k chronomere respectively
According to final to form the GPS sequence data that length is 3k.Stroke is defined as the stroke of k chronomere before the secondary stroke early period, in
Phase is defined as the stroke of most middle k chronomere, and the later period is defined as the stroke of last k chronomere.When journey time compared with
In short-term, it there may be intersection in three trip segments.When k value is larger, and journey time is shorter, it is understood that there may be take null value
Situation is filled using the null value for the mean value cross section for having value at this time.
For the speed of vehicle, the speed of preceding k chronomere or the speed of rear k chronomere can be intercepted as reference
Value；For the acceleration of vehicle, the acceleration of preceding k chronomere or the acceleration of rear k chronomere can be intercepted as ginseng
Examine value；For the average speed of vehicle, the average speed of preceding k chronomere or the average speed of rear k chronomere can be intercepted
Degree is used as reference value.Or use other known data cutout mode.
After completing feature construction step S2, the study of eXecute UML and training step S3, it is according to above-mentioned data set
A machine learning model is constructed and trained with driving habit feature.
Specifically, based on the data collection and driving habit feature construction machine learning model, in the following ways to mould
Type is learnt and is trained:
First the data set and driving habit feature are inputted in a GBDT model, take out the tree that the GBDT model generates
Index.
Simply introduce GBDT model herein, GBDT full name gradient decline tree, also referred to as MART, GBRT, Tree Net or
Treelink is to one of best several algorithms of true fitting of distribution inside conventional machines learning algorithm, and GBDT passes through
More wheel iteration, every wheel iteration generate a Weak Classifier, and each classifier is instructed on the basis of the residual error of last round of classifier
Practice.Requirement to Weak Classifier is usually simple enough, and is low variance and high deviation.Because the process of training is to pass through
Deviation is reduced the precision of final classification device is continuously improved.
Weak Classifier can generally be selected as CART TREE (namely post-class processing).Due to above-mentioned high deviation and simply
The depth of each post-class processing of requirement will not be very deep.Final total classifier is that the Weak Classifier for getting every training in rotation adds
(the namely addition model) that power summation obtains.
Then, the mode for reusing one-hot is input to a LR (Logistic Regression, logistic regression) model
In, recognition result is exported by the LR model.
The present embodiment by using GBDT+LR Fusion Model, can effective lifting system identification accuracy and reliability.
The update and deployment of the embodiment of the present invention model use following steps:
Docker mirror image is constructed on line, the model and prediction code are copied in the docker, when receiving
The history running data of the vehicle is inputted into the docker when history running data of vehicle, is handled by the docker,
Output in real time differentiates result and saves.
Docker is natural micro services, quick can efficiently solve many pain spots of deep learning, comprising:
Partial nerve network frame such as caffe dependence is overweight, difficult to install；It is excellent that various network models do not do engineering
Change, deployment is difficult；The frames such as Tensorflow are difficult to flexibly control, etc. to the occupancy of the hardware such as GPU.Therefore, the present embodiment
Machine learning model is updated and is disposed by the way of constructing docker mirror image on line, can not only evade the above problem,
And can simplify system, recognition speed is improved, accuracy of identification is improved.
Referring to shown in Fig. 2, the present embodiment correspondingly provides a kind of driver's replacement identification device 100, comprising:
Driving data obtains module 10, for acquiring the history running data of vehicle by OBD equipment；
Feature construction module 20, for constructing data set and driving habit feature；
The study of model and training module 30, for according to the data set and driving habit latent structure and training machine
Learning model；
The update and deployment module 40 of model, for the machine learning model to be updated and disposed.
Wherein, the feature construction module further comprises:
Data pre-processing unit 21, for being solved to the collected history running data of institute in driving data obtaining step
Code, cleaning, classification and storage, with excluding outlier and assigning null data；
Data set construction unit 22, for using pseudo label method building sample and data set；With
Driving habit feature construction unit 23, for based on stroke characteristic, three anxious features and GPS data building driver
Driving habit feature, the stroke characteristic include stroke time started, stroke end time, drive duration, mileage number, average speed
Degree and hypervelocity number；Described three anxious features include anxious acceleration, anxious deceleration and zig zag number and the time point of each run.
Further, the data set construction unit 22 is specifically used for:
Assuming that current vehicle driver does not replace, it, will when constructing sample using the history running data of current vehicle
It is set to 0 as negative sample, and by label；When extracting history running data building sample from other vehicles at random, made
For positive sample, and label is set to 1；
Then, all negative samples and positive sample are collected, forms the data set.It is easily understood that later use mould
When type verifies data set, when the sample label of a certain run-length data be 0 when, it is known that the driver of the vehicle there is no
Change, when the sample label of a certain run-length data is 1, it is known that the driver of the vehicle is changed.
In the present embodiment, when considering driving habit feature, mainly to stroke time started, stroke end time, starting
Point coordinate, terminal point coordinate, drive duration, mileage number, average speed, stroke maximum speed, stroke minimum speed, the secondary stroke away from
Divided from stroke residence time next time, maximum hypervelocity speed, hypervelocity number, the three anxious dimensions such as feature and GPS data
Analysis, wherein carrying out analysis in the data to above-mentioned each dimension is mainly to be handled GPS data and analyzed, comprising:
The elongated sequence data in the history running data is converted into fixed length sequence by the way of interception and filling
GPS data in each run is divided into three parts using block sampling method, intercepts K respectively from each part by data
A GPS point, the GPS sequence data that composition length is 3K；
Specifically, interception is divided into three parts, and stroke early, middle, late stage intercepts the number of strokes of k chronomere respectively
According to final to form the GPS sequence data that length is 3k.Stroke is defined as the stroke of k chronomere before the secondary stroke early period, in
Phase is defined as the stroke of most middle k chronomere, and the later period is defined as the stroke of last k chronomere.When journey time compared with
In short-term, it there may be intersection in three trip segments.When k value is larger, and journey time is shorter, it is understood that there may be take null value
Situation is filled using the null value for the mean value cross section for having value at this time.
For the speed of vehicle, the speed of preceding k chronomere or the speed of rear k chronomere can be intercepted as reference
Value；For the acceleration of vehicle, the acceleration of preceding k chronomere or the acceleration of rear k chronomere can be intercepted as ginseng
Examine value；For the average speed of vehicle, the average speed of preceding k chronomere or the average speed of rear k chronomere can be intercepted
Degree is used as reference value.Or use other known data cutout mode.
In conclusion replacing recognition methods compared to traditional driver, the present invention is by obtaining driving data, building number
It not only can protect the hidden of user without installing camera according to collection and driving habit feature, and construction machine learning model
It is private, additionally it is possible in conjunction with the external environment of vehicle history traveling record and current driving, accurately to analyze driving for current driver's
Behavioural habits are sailed, see whether match with history driving habit, to identify whether to replace driver.And by using GBDT+
The mode of LR Model Fusion effectively improves the stability and precision of model identification.
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Protect range.
Claims (10)
1. a kind of driver replaces recognition methods, which comprises the following steps:
Driving data obtaining step acquires the history running data of vehicle by OBD equipment；
Feature construction step constructs data set and driving habit feature；
The study of model and training step, according to the data set and driving habit latent structure and training machine learning model；
The update and deploying step of model.
2. driver as described in claim 1 replaces recognition methods, which is characterized in that the feature construction step is further wrapped
It includes:
Data prediction, the collected history running data of institute in driving data obtaining step is decoded, is cleaned, is classified and
Storage, with excluding outlier and assigning null data；
Data set building constructs sample and data set using pseudo label method；With
Driving habit feature construction, the driving habit feature based on stroke characteristic, three anxious features and GPS data building driver,
The stroke characteristic includes stroke time started, stroke end time, drives duration, mileage number, average speed and hypervelocity number；
Described three anxious features include anxious acceleration, anxious deceleration and zig zag number and the time point of each run.
3. driver as claimed in claim 2 replaces recognition methods, which is characterized in that described to construct sample using pseudo label method
And the step of data set, specifically includes:
Assuming that current vehicle driver does not replace, when constructing sample using the history running data of current vehicle, made
For negative sample, and label is set to 0；When extracting history running data building sample from other vehicles at random, as just
Sample, and label is set to 1；
After all negative samples and positive sample are collected, the data set is formed.
4. driver as claimed in claim 2 replaces recognition methods, which is characterized in that the step of the driving habit feature construction
Suddenly it specifically includes:
The elongated sequence data in the history running data is converted into fixed length sequence data by the way of interception and filling,
The GPS data in each run is divided by three parts using block sampling method, intercepts K GPS respectively from each part
Point, the GPS sequence data that composition length is 3K；
If journey time is shorter or K value is larger causes data to there is intersection, using the mean value for having value to cross section
Null value is filled.
5. driver as described in claim 1 replaces recognition methods, which is characterized in that the study of the model and training step
It specifically includes:
Collection and driving habit feature construction machine learning model based on the data, and model is learnt in the following ways
And training:
First the data set and driving habit feature are inputted in a GBDT model, take out the tree rope that the GBDT model generates
Draw；
The mode for reusing one-hot is input in a LR model, exports recognition result by the LR model.
6. driver as described in claim 1 replaces recognition methods, which is characterized in that the update and deploying step of the model
It specifically includes:
Docker mirror image is constructed on line, the model and prediction code are copied in the docker, when receiving vehicle
History running data when the history running data of the vehicle inputted into the docker, handled by the docker, in real time
Output differentiates result and saves.
7. a kind of driver replaces identification device characterized by comprising
Driving data obtains module, for acquiring the history running data of vehicle by OBD equipment；
Feature construction module, for constructing data set and driving habit feature；
The study of model and training module, for learning mould according to the data set and driving habit latent structure and training machine
Type；
The update and deployment module of model, for the machine learning model to be updated and disposed.
8. driver as claimed in claim 7 replaces identification device, which is characterized in that the feature construction module is further wrapped
It includes:
Data pre-processing unit, for being decoded, clearly to the collected history running data of institute in driving data obtaining step
It washes, classify and stores, with excluding outlier and assigning null data；
Data set construction unit, for using pseudo label method building sample and data set；With
Driving habit feature construction unit is practised for the driving based on stroke characteristic, three anxious features and GPS data building driver
Used feature, the stroke characteristic include stroke time started, stroke end time, drive duration, mileage number, average speed and surpass
Fast number；Described three anxious features include anxious acceleration, anxious deceleration and zig zag number and the time point of each run.
9. driver as claimed in claim 8 replaces identification device, which is characterized in that the data set construction unit is specifically used
In:
Assuming that current vehicle driver does not replace, when constructing sample using the history running data of current vehicle, made
For negative sample, and label is set to 0；When extracting history running data building sample from other vehicles at random, as just
Sample, and label is set to 1；
After all negative samples and positive sample are collected, the data set is formed.
10. driver as claimed in claim 8 replaces identification device, which is characterized in that driving habit feature construction unit tool
Body is used for,
The elongated sequence data in the history running data is converted into fixed length sequence data by the way of interception and filling,
The GPS data in each run is divided by three parts using block sampling method, intercepts K GPS respectively from each part
Point, the GPS sequence data that composition length is 3K；
If journey time is shorter or K value is larger causes data to there is intersection, using the mean value for having value to cross section
Null value is filled.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910698186.5A CN110458214B (en) | 2019-07-31 | 2019-07-31 | Driver replacement recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910698186.5A CN110458214B (en) | 2019-07-31 | 2019-07-31 | Driver replacement recognition method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110458214A true CN110458214A (en) | 2019-11-15 |
| CN110458214B CN110458214B (en) | 2023-07-07 |
Family
ID=68484078
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910698186.5A Active CN110458214B (en) | 2019-07-31 | 2019-07-31 | Driver replacement recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110458214B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814667A (en) * | 2020-07-08 | 2020-10-23 | 山东浪潮云服务信息科技有限公司 | Intelligent road condition identification method |
| CN112918472A (en) * | 2019-12-05 | 2021-06-08 | 奥迪股份公司 | Vehicle driving assistance system, vehicle using the same, and corresponding method and medium |
| CN113361412A (en) * | 2021-06-08 | 2021-09-07 | 西南科技大学 | Finger vein image accurate matching method based on ULBP and SURF feature progression |
| CN114999022A (en) * | 2022-05-19 | 2022-09-02 | 成都亿盟恒信科技有限公司 | Driving habit analysis method and system based on historical driving data |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197633A (en) * | 2017-11-24 | 2018-06-22 | 百年金海科技有限公司 | Deep learning image classification based on TensorFlow is with applying dispositions method |
| WO2018139531A1 (en) * | 2017-01-26 | 2018-08-02 | パナソニックＩｐマネジメント株式会社 | Information processing system, information processing method, program, and vehicle |
| CN108537198A (en) * | 2018-04-18 | 2018-09-14 | 济南浪潮高新科技投资发展有限公司 | A kind of analysis method of the driving habit based on artificial intelligence |
| CN108995653A (en) * | 2018-07-06 | 2018-12-14 | 北京理工大学 | A kind of driver's driving style recognition methods and system |
| CN109118055A (en) * | 2018-07-19 | 2019-01-01 | 众安信息技术服务有限公司 | A kind of driving behavior methods of marking and device |
| CN109948668A (en) * | 2019-03-01 | 2019-06-28 | 成都新希望金融信息有限公司 | A kind of multi-model fusion method |
-
2019
- 2019-07-31 CN CN201910698186.5A patent/CN110458214B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018139531A1 (en) * | 2017-01-26 | 2018-08-02 | パナソニックＩｐマネジメント株式会社 | Information processing system, information processing method, program, and vehicle |
| CN108197633A (en) * | 2017-11-24 | 2018-06-22 | 百年金海科技有限公司 | Deep learning image classification based on TensorFlow is with applying dispositions method |
| CN108537198A (en) * | 2018-04-18 | 2018-09-14 | 济南浪潮高新科技投资发展有限公司 | A kind of analysis method of the driving habit based on artificial intelligence |
| CN108995653A (en) * | 2018-07-06 | 2018-12-14 | 北京理工大学 | A kind of driver's driving style recognition methods and system |
| CN109118055A (en) * | 2018-07-19 | 2019-01-01 | 众安信息技术服务有限公司 | A kind of driving behavior methods of marking and device |
| CN109948668A (en) * | 2019-03-01 | 2019-06-28 | 成都新希望金融信息有限公司 | A kind of multi-model fusion method |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112918472A (en) * | 2019-12-05 | 2021-06-08 | 奥迪股份公司 | Vehicle driving assistance system, vehicle using the same, and corresponding method and medium |
| CN111814667A (en) * | 2020-07-08 | 2020-10-23 | 山东浪潮云服务信息科技有限公司 | Intelligent road condition identification method |
| CN111814667B (en) * | 2020-07-08 | 2022-10-14 | 山东浪潮云服务信息科技有限公司 | Intelligent road condition identification method |
| CN113361412A (en) * | 2021-06-08 | 2021-09-07 | 西南科技大学 | Finger vein image accurate matching method based on ULBP and SURF feature progression |
| CN114999022A (en) * | 2022-05-19 | 2022-09-02 | 成都亿盟恒信科技有限公司 | Driving habit analysis method and system based on historical driving data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110458214B (en) | 2023-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110458214A (en) | Driver replaces recognition methods and device | |
| Dong et al. | Characterizing driving styles with deep learning | |
| CN107492251B (en) | Driver identity recognition and driving state monitoring method based on machine learning and deep learning | |
| CN111179585B (en) | Site testing method and device for automatic driving vehicle | |
| CN110329271A (en) | A kind of multisensor vehicle driving detection system and method based on machine learning | |
| JP2020530578A (en) | Driving behavior scoring method and equipment | |
| CN110874362A (en) | Data association analysis method and device | |
| DE102016210453A1 (en) | Vehicle, system in communication with a communication module of the vehicle, and system in communication with a group of vehicles | |
| CN110395245B (en) | Hybrid electric vehicle energy management system based on fixed route driving information | |
| CN114450557B (en) | Route deviation quantification and vehicle route learning based thereon | |
| CN110727706A (en) | Method for rapidly extracting and grading risk driving scene for intelligent networking automobile test | |
| CN109993966A (en) | A kind of method and device of building user portrait | |
| CN115291515A (en) | Automatic driving simulation test system and evaluation method based on digital twinning | |
| CN115797403A (en) | Traffic accident prediction method and device, storage medium and electronic device | |
| Chu et al. | A review of driving style recognition methods from short-term and long-term perspectives | |
| US20210374598A1 (en) | A System and Method for Using Knowledge Gathered by a Vehicle | |
| KR102462403B1 (en) | Lightweight Driver Behavior Identification Model with Sparse Learning on In-vehicle CAN-BUS Sensor Data | |
| CN111027859B (en) | Driving risk prevention method and system based on motor vehicle state monitoring data mining | |
| CN112269827A (en) | Data processing method and device, computer equipment and computer readable storage medium | |
| CN116340332A (en) | Method and device for updating scene library of vehicle-mounted intelligent system and vehicle | |
| US20230391357A1 (en) | Methods and apparatus for natural language based scenario discovery to train a machine learning model for a driving system | |
| CN108460057B (en) | User travel mining method and device based on unsupervised learning | |
| CN112308434A (en) | Traffic safety risk assessment method and system | |
| CN106097751A (en) | Vehicle travel control method and device | |
| CN112308171A (en) | Vehicle position prediction modeling method based on simulated driver |
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 |