CN110276563A - A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model - Google Patents
A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model Download PDFInfo
- Publication number
- CN110276563A CN110276563A CN201910583174.8A CN201910583174A CN110276563A CN 110276563 A CN110276563 A CN 110276563A CN 201910583174 A CN201910583174 A CN 201910583174A CN 110276563 A CN110276563 A CN 110276563A
- Authority
- CN
- China
- Prior art keywords
- transfer
- data
- point
- seconds
- mode
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012546 transfer Methods 0.000 title claims abstract description 113
- 238000000034 method Methods 0.000 title claims abstract description 47
- 239000013598 vector Substances 0.000 title claims abstract description 38
- 230000000694 effects Effects 0.000 title claims abstract description 34
- 230000006399 behavior Effects 0.000 claims abstract description 25
- 230000000284 resting effect Effects 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims abstract description 6
- 230000001133 acceleration Effects 0.000 claims description 16
- 230000008859 change Effects 0.000 claims description 11
- 230000003542 behavioural effect Effects 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 7
- 238000012706 support-vector machine Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 4
- 241001269238 Data Species 0.000 claims description 3
- 230000008901 benefit Effects 0.000 claims description 3
- 230000008092 positive effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000000295 complement effect Effects 0.000 description 3
- 238000011835 investigation Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000013456 study Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 241000976416 Isatis tinctoria subsp. canescens Species 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012856 packing Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000153 supplemental effect Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- 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
- G06Q50/40—Business processes related to the transportation industry
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Data Mining & Analysis (AREA)
- Educational Administration (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Navigation (AREA)
Abstract
The mode of transportation that the invention discloses a kind of based on supporting vector machine model changes to Activity recognition method, includes the following steps: Step 1: basic data acquisition and constructing data base management system;Step 2: being pre-processed to basic data;Step 3: being identified to the resting state of individual trip;Step 4: calculating the characteristic index of different mode of transportation transfer behaviors;Step 5: building supporting vector machine model carries out transfer Activity recognition to overall process trip data;Step 6: the relevant information of transfer behavior is extracted in matching from data base management system.Compared with prior art, the positive effect of the present invention is: the method for the present invention can be used in identification and extract the information such as time, place, the number of Individual Mode transfer, can be provided strong support to promote China's traffic transfer information collection content and data precision, can be used for large sample, wide region, high-precision, automation traffic trip transfer information collection.
Description
Technical field
The invention belongs to Intelligent traffic information acquiring fields more particularly to a kind of utilization supporting vector machine model to satellite
Position data are analyzed, are pre-processed, model construction and application, thus the whole all modes of transportation transfers of automatic identification individual trip
The techniqueflow and method of behavior.
Background technique
The fast development in city brings an infinite variable and extremely complex multidimensional Traffic Systems, huge solid
Facility network, polynary synthesis trip mode, the trip purpose of sequential correlation, city individual daily trip feature is increasingly difficult to
To predict and capture.Mode of transportation transfer information is the important content of individual trip information acquisition, can be different for objective assurance
Mode of transportation trip requirements, the reasonable traffic programme of guidance, optimization traffic infrastructure etc. play an important role and city intelligent
The significant data basis of traffic information system building.
It is existing that about mode of transportation transfer information collection, there are mainly two types of methods:
(1) papery questionnaire: papery questionnaire method refers to by formulating questionnaire careful in detail, it is desirable that surveyee is accordingly
Answered the method with data collection.When carrying out mode of transportation transfer behavior acquisition using papery questionnaire, surveyee needs
The information such as the continuous whole transfer times recalled during one day or multiple days is gone on a journey, transfer manner, transfer site, it is negative by investigation
Misgivings in terms of load and individual privacy, papery questionnaire have the following disadvantages a little: 1) data deviation, mistake caused by subjective memory
Phenomenon is universal.It being influenced by the subjective wish that participates in of surveyee, trip memory deviation, mistake, random phenomenon are universal, particularly with
Short time and short distance trip, information are omitted obvious;2) research cost is high, tissue difficulty is big.The other resident trip of general City-level
Investigation needs to set up special leading group, coordinates multi-sector cooperation, training door-to-door survey, it is larger and some to organize and implement difficulty
Big city comprehensive transport plan establishment in be only used for traffic study expense be as high as it is up to ten million;3) data dynamic is poor, updates
Period is long.Since at high cost, enforcement difficulty is big, large-scale resident trip survey is often just carried out once every 5-10, north
The megalopolis such as capital, Shanghai, Shenzhen were also just implemented 5-6 times since the establishment of the nation, and considerable tier 2 cities may be not carried out so far
It crosses primary.Under the background that Urbanization in China is accelerated, transport need acutely expands, papery questionnaire data is difficult to reflect in time
The true transport need of rapid urban formulates traffic programme scheme by stale data and is necessarily difficult to play substantial effect.
(2) internet writing-method: with intelligent chip technology, the fast development of Internet technology and universal, current resident
Taking the modes such as subway, bus, shared bicycle can be paid by individual handset or transportation card, these intelligently pay
System, such as shared bicycle APP, subway IC card, when can quickly identify userspersonal information, locating geographical location, trip
Between etc. information, thus indirectly reflection mode of transportation change to behavior, relative to traditional questionnaire method, data precision is obtained
It is good to be promoted.However, there is also certain technological deficiencies, such as bus IC card to swipe the card for such technology, it is merely able to obtain and gets on the bus ground
Point information can not but record location information of getting off, and transfer point, which exists, omits phenomenon;In addition, car, taxi, walking etc. are made
For important daily mode of transportation, internet writing-method can not also obtain the transfer information of these trip modes at present.
Therefore, from the point of view of Current traffic mode changes to information identification predicament, it would be highly desirable to seek a kind of more intelligent, pervasive, smart
Substitution or promotion of the true technological means as the prior art.
In recent years, global navigational satellite positioning system (GNSS) is fast-developing, in addition to the more GPS of current application, I
State's BEI-DOU position system, Russian Glonass global position system and European Galileo satellite positioning system etc. are all fast-developing
And application is put into, a large amount of objective, full and accurate, dynamic positioning track data are provided for individual travel behaviour analysis.At the same time,
Smart phone, wearable device (motion bracelet, wrist-watch etc.) etc. are quickly popularized with the vehicle equipment of satellite positioning chip, to defend
The a wide range of of star location data, large sample acquisition provide unprecedented opportunity, also change for the mode of transportation based on GNSS location data
Multiply Activity recognition and provides good opportunity.Therefore the present invention is attempted by being analyzed GNSS location data, being pre-processed, algorithm
Building and application extract transfer time, transfer site, number of transfer etc. so that automatic identification individual traffic mode changes to behavior
Details.The technology relative to existing acquisition method, have sample size is bigger, coverage area is wider, real time and dynamic is more preferable,
The more high many advantages of data precision.
Summary of the invention
In order to overcome the disadvantages mentioned above of the prior art, the invention proposes a kind of traffic sides based on supporting vector machine model
Formula changes to Activity recognition method, for the existing mode of transportation transfer Activity recognition method somewhat expensive in China, organizes cumbersome, data
The technological deficiencies such as coarse are proposed a kind of based on support vector machines mould by mining analysis individual trip satellite location data feature
The mode of transportation of type changes to behavior intelligent identification technology.Firstly, analysis individual trip GNSS satellite location data space-time characteristic, knows
Not individual moving condition;Secondly, analyzing different mode of transportation transfer Behavior laws, refines mode of transportation and change to behavior representation parameter
And index value;Finally, the input parameter based on selection, specific aim constructs supporting vector machine model and carries out the knowledge of mode of transportation transfer point
Not, and matching is indexed from raw data base extract the details such as transfer time, transfer site.This method advantage is: sufficiently
The high-precision of GNSS satellite location data is utilized, can continuously track the feature of trip track, and plays supporting vector machine model
Excellent study and intelligent recognition ability, realize using satellite location data carry out resident's mode of transportation transfer behavior intelligence knowledge
Not.The invention can be used for large sample, wide region, high-precision, automation traffic trip transfer information collection.
The technical solution adopted by the present invention to solve the technical problems is: a kind of traffic side based on supporting vector machine model
Formula changes to Activity recognition method, includes the following steps:
Step 1: basic data acquisition and constructing data base management system;
Step 2: being pre-processed to basic data;
Step 3: being identified to the resting state of individual trip;
Step 4: calculating the characteristic index of different mode of transportation transfer behaviors;
Step 5: building supporting vector machine model carries out transfer Activity recognition to overall process trip data;
Step 6: the relevant information of transfer behavior is extracted in matching from data base management system.
Compared with prior art, the positive effect of the present invention is:
The present invention innovatively proposes to be analyzed and processed satellite positioning track using supporting vector machine model, to know
Other individual traffic mode changes to the entire work flow and method of behavior.Specifically include data acquisition and database construction method, data
Preprocess method, trip state identification method, different transfer behavioural characteristic index selections and setting method, supporting vector machine model
Building and transfer Activity recognition application method etc..This method can be used in identification extract Individual Mode transfer when
Between, place, the information such as number.The present invention can provide effectively to promote China's traffic transfer information collection content and data precision
It supports.
Detailed description of the invention
Examples of the present invention will be described by way of reference to the accompanying drawings, in which:
Fig. 1 is individual trip satellite positioning track space-time characteristic schematic diagram;
Fig. 2 is that support vector machines identifies that mode of transportation changes to behavior schematic diagram;
Fig. 3 is constructed data base management system interface;
Fig. 4 wades out capable satellite location data sample for one;
Fig. 5 is that positioning track repairs result schematic diagram;
Fig. 6 is dwell point recognition result schematic diagram;
Fig. 7 is dwell point Statistical error result schematic diagram;
Fig. 8 is final transfer point recognition result schematic diagram.
Specific embodiment
A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model, first with GNSS satellite
The portable terminal (such as smart phone, bracelet) of positioning chip acquire individual whole day satellite positioning track data (including
Subscriber-coded, travel time, longitude and latitude data, speed etc.), on the basis of data prediction, known using anchor point time and space idea
Other individual movement state (movement or static);Secondly different mode of transportation transfer behaviors front and backs data rule and feature are analysed in depth
Otherness determines and calculates acquisition transfer behavioural characteristic indicators, changes to Activity recognition for mode of transportation and provide mode input
Basis;Finally, building supporting vector machine model carries out transfer point identification to overall process trip data, and accordingly from raw data base
The information such as transfer time, transfer site are extracted in middle matching.
Detailed description are as follows by the following examples and in conjunction with attached drawing to the method for the present invention:
It is as shown in Figure 1 individual trip satellite positioning track space-time characteristic schematic diagram.During individual trip, satellite positioning
Technology can continuously get record trip track ready, and the density degree of the locus of points is capable of dynamic/quiet state of well-characterized individual activity
Feature.As shown in Figure 1, in two-dimensional surface packing phenomenon occurs for tracing point when individual trip is stopped, three-dimensional space then at any time
Between extend vertically upwards;When individual movement, the locus of points is got ready in 2D and the equal regularity of 3d space, and point spacing has with trip speed
It closes, more spacing is bigger as soon as possible for speed.Therefore between the density in satellite site, point the parameters such as distance, travel time in certain journey
Individual movement state has been reacted on degree, has been excavated by the law-analysing and internal relation of tracing point space-time characteristic parameter, Neng Gougao
Effect identification individual movement and stationary motion state.
It is illustrated in figure 2 support vector machines identification mode of transportation transfer behavior schematic diagram.Wherein left figure is characterized parameter and exists
The display of lower dimensional space is as a result, data mixing phenomenon is serious, it is not easy to two category feature data of cutting;Right figure is data in higher-dimension sky
Between mapping result, it can be seen that become to divide in higher-dimension in the data of higher dimensional space, script linearly inseparable.Supporting vector
The excellent ability of machine model is to different be mapped to higher-dimension by a variety of using mapping function wait distinguishing transfer characteristic parameter and put down
Face, so that the inseparable characteristic parameter of script low-dimensional is split identification in higher dimensional space, mathematics essence is to solve for mapping
Optimal convex programming equation on feature space afterwards.
The introduction of traffic transfer information extraction technology is carried out by taking certain one day trip data of individual as an example:
Step 1: basic data acquisition and data base management system construct
1) basic data acquisition: using the portable data gathering equipment with GNSS satellite positioning function (such as intelligent hand
Machine, motion bracelet, professional satellite positioning device etc.) acquisition individual trip track data, including it is the subscriber-coded, travel time, fixed
The contents such as site longitude and latitude.In data acquisition, it is desirable that experimenter carries data acquisition equipment at any time, and data sampling frequency pushes away
It recommends as 1 second every time.Investigation terminates, and collected data are uploaded to computer data base management system.
Different depending on the application, the basic data in database is divided into two major classes.The first kind is model training data, that is, is used for
The parameter configuration and debugging of supporting vector machine model, training data should change to behavioral data comprising all common modes of transportation, such as
Bicycle is changed in walking, public transport is changed in walking, car etc. is changed in walking, and in order to reach training effect, all kinds of transfer behavioral datas should not
It less than 100 groups, and should include the data that congestion period and non-congestion period acquire;Second class is target data to be identified, is used for
It inputs trained model and carries out traffic transfer Activity recognition.
When the present embodiment carries out basic data acquisition, front opening mobile satellite location equipment (this example use of going out of getting up individual morning
Smart phone acquires GPS positioning data) prepare data acquisition, equipment is carried after GPS positioning is stablized and is normally gone on a journey, and is gone forward side by side
Row daily routines in one day, such as working, meeting good friend.When individual come home from work decide not to trip after close software, and will
A whole day data collected are uploaded to data base management system.Individual whole day trip GPS track data include it is subscriber-coded, go out
The contents such as row time, anchor point longitude and latitude, data acquiring frequency are recommended as 1 second.Data sample is as shown in table 1.
1 GPS satellite location data content of table and format
2) data base management system constructs: constructing oracle database using SaaS framework, software meets user's customization
Data acquisition demand and multi-user's Parallel Service, database have according to time and subscriber-coded real-time query and export function,
And guarantees data efficient storage and read.It is illustrated in figure 3 constructed data base management system interface.
Step 2: data prediction
Due to GNSS satellite signal by physical shielding indoors or when underground is gone on a journey, and data positioning accuracy is by weather, temperature
The factors such as degree, outside environment influence.Therefore it needs to pre-process original trip track data.One wades out capable satellite positioning
Data sample is as shown in Figure 4, it can be seen that trip has signal deletion section on the way, needs to carry out data reparation.
(1) missing data is filled up: according to equipment location frequency, when successively searching for shortage of data section during going on a journey and carrying out
Sequence label.For each missing track, uniform interpolation complement point is carried out using anchor point longitude and latitude before and after missing section and positioning time,
Data filling frequency is consistent with data acquiring frequency.The complement point frequency of the present embodiment is 1 second each.
(2) wrong data identification and amendment: the whole front and back positioning space of anchor point longitude and latitude Continuous plus trip is utilized
From removal point spacing is greater than 50 meters of anchor points, and (common urban transit system mode moving distance per second is greater than 50 meters of positioning less than 50 meters
Point is usually that satellite-signal fluctuation generates), and according to step (1) missing data complementing method interpolation complement point again.Data
Pre-processed results are as shown in figure 5, draw circle part wherein as the data point after reparation.
Step 3: trip resting state identification
When mode of transportation transfer occurs for individual, motion state is usually static.The density in satellite site, point
Between the information such as distance, travel time reacted individual movement state to a certain extent, can be used in identification individual in movement
Or resting state.
(1) trip stops identification: with 3 minutes for time threshold, 70 meters are distance threshold, and 150 points are anchor point quantity threshold
Value carries out stay segment identification, when each orbit segment is when 3 minutes, 70 meters of default number of sites of range are more than at 150, then it is assumed that individual exists
Position section is stopped, and is merged to the stay segment for continuously having overlapping, and carries out timing label to stay segment.Dwell point
Recognition result as shown in fig. 6, wherein this trip identify 1 to 24 place altogether, totally 24 dwell points.
(2) recognition result optimizes: due to stopping in dwell point comprising stop (stopping to trip purpose) and short time for a long time
It stays (transfer point stops, signal lamp stops etc.), the removal of this optimization order stops identification error for a long time, and signal lamp stops error will
It removes in steps of 5.With 15 minutes for time threshold, 200 meters are distance threshold, and 700 points are stopped for anchor point amount threshold
Section is stayed to identify, when each orbit segment is when 15 minutes, 200 meters of default number of sites of range are more than at 700, then it is assumed that individual is in the position
Section has occurred stops (non-transfer in short-term stops) for a long time.Identification stops segment number and removes stop record.Removal is for a long time
Optimum results after stop as shown in fig. 7, wherein 1,8,13,24 be optimization removal long-time dwell point, i.e. destination stops,
Remaining is the short time dwell point identified.
Step 4: difference transfer behavioural characteristic setup measures:
When different modes of transportation are changed to, transfer time length, transfer front and back speed, transfer fore-aft acceleration etc. exist bright
Significant difference is different, therefore different mode of transportation transfer behavioural characteristic index selections and calculation method are as follows:
(1) points are stopped: stopping segment number according in trip resting state identification, the stay segment last bit point time is subtracted and is stopped
A section starting time is stayed, time difference length transition is number of seconds, as stops points.It successively calculates each transfer section and stops points.
(2) it changes to preceding 60 seconds speed: on the basis of each stay segment starting point, extending 60 seconds forward, utilize continuous two o'clock
It is displaced between calculation of longitude & latitude point, calculates the time difference using the continuous two-point locating time, thus before displacement determines transfer divided by duration
60 seconds each anchor point instantaneous velocitys.
(3) 60 seconds speed after changing to: on the basis of each stay segment end point, extend 60 seconds backward, utilize continuous two o'clock
It is displaced between calculation of longitude & latitude point, calculates the time difference using the continuous two-point locating time, thus after displacement determines transfer divided by duration
60 seconds each anchor point instantaneous velocitys.
(4) preceding 60 seconds acceleration are changed to: on the basis of each stay segment starting point, extending 60 seconds forward, utilizes continuous two o'clock
Instantaneous velocity values calculating speed it is poor, using the continuous two-point locating time calculate the time difference, so that speed difference is true divided by the time difference
First 60 seconds each anchor point acceleration values of transfer calmly.
(5) 60 seconds acceleration after changing to: on the basis of each stay segment end point, extend 60 seconds backward, utilize continuous two o'clock
Instantaneous velocity values calculating speed it is poor, using the continuous two-point locating time calculate the time difference, so that speed difference is true divided by the time difference
60 seconds each anchor point acceleration values after fixed transfer.
Step 5: supporting vector machine model building and transfer Activity recognition
When the different vehicles are changed to, walking is usually the transient mode of other vehicles, therefore, is compiled based on Matlab
Journey constructs supporting vector machine model to identify 6 kinds of common mode of transportation transfer behaviors, including walking transfer bicycle, walking transfer
Bus, walking transfer car and bicycle transfer walking, bus transfering walking and car change to walking.
Since the basic principle of algorithm of support vector machine is that lower dimensional space linearly inseparable data are mapped to higher dimensional space,
With target classification best (spacing is maximum) for target, the optimum linearity classifier on feature space is found, can finally be converted into one
The solution of a convex quadratic programming problem.But when low-dimensional data in practice, being mapped to higher dimensional space solution convex programming equation, to
The problem of amount mapping usually will appear dimension explosion and can not solve, therefore support vector machines introduces kernel function concept.It will calculate
Interior Product function of two vectors after the implicit mapping in space is called kernel function, replaces the direct inner product of original higher dimensional space with this
Accounting equation is avoided and is directly calculated in higher dimensional space, and result is of equal value.Gaussian kernel function is to generally acknowledge at present
The kernel function of better performances.Gaussian kernel function has a nuclear parameter σ to need to demarcate, if nuclear parameter σ selects very big, high order feature
On weight decay very fast, so being effectively equivalent to the subspace of a low-dimensional;If σ selects to obtain very little, can incite somebody to action
Arbitrary data are mapped as linear separability, may cause serious overfitting problem.
In addition, a small amount of noise data can largely influence the selection of best hyperplane in practice.Therefore, in order to balance
The influence of maximum classifying distance and error information introduces penalty coefficient C.Penalty coefficient C can be in determining proper subspace
The ratio of Learning machine fiducial range and empiric risk is adjusted so that the Generalization Ability of Learning machine is best.
Therefore, best in order to reach mode of transportation transfer point recognition effect, the present invention is established using Matlab programming and is supported
Vector machine model, chooses mapping function of the gaussian kernel function as model, and the nuclear parameter of supporting vector machine model is set as 3, punishes
Penalty factor is set as 200.Input feature value is set as stop above-mentioned and counts, changes to 60 seconds speed after preceding speed, transfer in 60 seconds
60 seconds acceleration five indices after degree, transfer preceding acceleration, transfer in 60 seconds, output vector are 6 kinds of transfer modes.The step can
Effective district sub-signal lamp stops and transfer behavior stops, to identify extraction transfer information.Recognition result is as shown in figure 8, in figure
1-8 at for eliminate signal lamp stop misrecognition point after 8 transfer points, indicate the sample trip altogether comprising 8 transfers row
For.
The present invention is as follows using the principle that 5 parameters described in step 4 carry out the identification of 6 kinds of mode of transportation transfer points:
It is one before every kind of transfer point for 6 kinds of different transfers by step 4 it is found that transfer phenomenon has occurred before and after dwell point
Mode of transportation is planted, is a kind of mode of transportation after transfer point, trip speed, acceleration of front and rear mode etc. (i.e. 5 features) have respectively
From the feature of mode, numberical range and rule are different.During the calibration process, by inputting 5 supplemental characteristics and demarcating
Known transfer manner, algorithm of support vector machine can learn automatically out different modes of transportation for the feature and rule of data.
When new data input trained supporting vector machine model, it will be able to identify the transfer having occurred between any two ways.
Step 6: transfer information matches are extracted
It is index with the starting point time for the behavior of changing to, from original number on the basis of mode of transportation changes to Activity recognition
Transfer point latitude and longitude information, transfer people's number information are extracted according to matching in library.It is the first sequence index with individual's number, with transfer
Time of origin is the second sequence index, and ascending order arranges all transfer data, that is, obtain individual one day continuous transfer time, place,
The traffic studys information needed such as number.Information sample such as 2 institutes of table such as final sample mode of transportation transfer time, place, the number
Show.
2 mode of transportation of table changes to information recognition result sample
We have carried out mass data acquisition in certain provincial capital, city, downtown and have tested with technical application.For walking-
It is a variety of before bicycle-walking, walking-bus-walking, walking-car-walking and walking, public transport, car
Combination transfer behavior (such as walking-public transport-walking-car-walking) has carried out 150 groups of tests respectively, and overall test data are adopted
Collecting duration is more than 260 hours, includes 936000 location informations.When data acquisition time covers the trip of early evening peak peace peak
Between.Test result shows that the transfer Activity recognition technology of the mode of transportation based on supporting vector model can be effective for traffic side
Formula changes to Activity recognition, and mode of transportation transfer recognition accuracy reaches 87%, and leakage discrimination control is within 10%;Change to point
Identification error control is set within 60 meters, transfer time identification error controlled within 3 minutes.This is relative to traditional questionnaire survey
For method (recall and forget serious, time error up to dozens of minutes), mode of transportation transfer information accuracy of identification has been obtained greatly
Width is promoted, and can well be used for traffic survey data acquisition practice.
Claims (10)
1. a kind of mode of transportation based on supporting vector machine model changes to Activity recognition method, it is characterised in that: including walking as follows
It is rapid:
Step 1: basic data acquisition and constructing data base management system;
Step 2: being pre-processed to basic data;
Step 3: being identified to the resting state of individual trip;
Step 4: calculating the characteristic index of different mode of transportation transfer behaviors;
Step 5: building supporting vector machine model carries out transfer Activity recognition to overall process trip data;
Step 6: the relevant information of transfer behavior is extracted in matching from data base management system.
2. a kind of mode of transportation based on supporting vector machine model according to claim 1 changes to Activity recognition method,
Be characterized in that: the basic data is the individual whole day satellite of the portable terminal acquisition with GNSS satellite positioning chip
Positioning track data, comprising: subscriber-coded, travel time, longitude and latitude data and speed.
3. a kind of mode of transportation based on supporting vector machine model according to claim 2 changes to Activity recognition method,
Be characterized in that: the basic data is divided into training data and data to be identified, and the training data is used for supporting vector machine model
Parameter configuration and debugging, change to behavioral datas comprising all common modes of transportation, all kinds of transfer behavioral datas are more than or equal to 100
Group, and the data including congestion period and the acquisition of non-congestion period;The data to be identified, for input it is trained support to
Amount machine model carries out traffic transfer Activity recognition.
4. a kind of mode of transportation based on supporting vector machine model according to claim 1 changes to Activity recognition method,
Be characterized in that: carrying out pretreated method to basic data described in step 2 includes being filled up to missing data and to error number
According to being modified, in which:
(1) missing data is filled up: according to equipment location frequency, successively searches for shortage of data section during going on a journey and carry out
Timing label;For each missing track, uniform interpolation benefit is carried out using anchor point longitude and latitude before and after missing section and positioning time
Point, data filling frequency are consistent with data acquiring frequency;
(2) wrong data is modified: goes on a journey whole front and back positioning space from going using anchor point longitude and latitude Continuous plus
Except spacing is greater than the wrong data of 50 meters of anchor points, and interpolation is mended again according to the missing data complementing method of (1) step
Point.
5. a kind of mode of transportation based on supporting vector machine model according to claim 1 changes to Activity recognition method,
It is characterized in that: the resting state of individual trip being carried out knowing method for distinguishing including following content described in step 3:
(1) trip dwell point is identified: will is more than 150 points of orbit segment identification in 3 minutes, 70 meters of default number of sites of range
For stay segment, the stay segment for continuously having overlapping is merged, and timing label is carried out to stay segment, obtains dwell point;
(2) dwell point is optimized, obtains short time dwell point: will be more than in 15 minutes, 200 meters of default number of sites of range
700 points of orbit segment is identified as long-time dwell point, and removing the dwell point obtained after long-time dwell point is to stop the short time
Point.
6. a kind of mode of transportation based on supporting vector machine model according to claim 5 changes to Activity recognition method,
Be characterized in that: difference mode of transportation described in step 4 change to behavior characteristic index include stop points, the preceding 60 seconds speed of transfer,
60 seconds acceleration after 60 seconds acceleration and transfer before speed, transfer in 60 seconds after transfer, calculation method is as follows:
(1) points are stopped: according to segment number is stopped, the stay segment last bit point time being subtracted into stay segment starting time, time difference length
Number of seconds is converted to, points are as stopped;It successively calculates each transfer section and stops points;
(2) it changes to preceding 60 seconds speed: on the basis of each stay segment starting point, extending 60 seconds forward, utilize the longitude and latitude of continuous two o'clock
Degree is displaced between calculating point, calculates the time difference using the continuous two-point locating time, so that displacement determines first 60 seconds of transfer divided by duration
Each anchor point instantaneous velocity;
(3) 60 seconds speed after changing to: on the basis of each stay segment end point, extend 60 seconds backward, utilize the longitude and latitude of continuous two o'clock
It is displaced between degree calculating point, calculates the time difference using the continuous two-point locating time, so that displacement determines after changing to 60 seconds divided by duration
Each anchor point instantaneous velocity;
(4) preceding 60 seconds acceleration are changed to: on the basis of each stay segment starting point, extending 60 seconds forward, utilizes the wink of continuous two o'clock
When velocity amplitude calculating speed it is poor, using the continuous two-point locating time calculate the time difference, thus speed difference divided by the time difference determination change
Multiply first 60 seconds each anchor point acceleration values;
(5) 60 seconds acceleration after changing to: on the basis of each stay segment end point, extend 60 seconds backward, utilize the wink of continuous two o'clock
When velocity amplitude calculating speed it is poor, using the continuous two-point locating time calculate the time difference, thus speed difference divided by the time difference determination change
60 seconds each anchor point acceleration values after multiplying.
7. a kind of mode of transportation based on supporting vector machine model according to claim 6 changes to Activity recognition method,
It is characterized in that: constructing the method for supporting vector machine model described in step 5 are as follows: establish support vector machines mould using Matlab programming
Type, in which: mapping function is gaussian kernel function, and nuclear parameter σ is 3, and penalty coefficient C is 200, and input feature value is dwell point
60 seconds acceleration, output vector are after acceleration, transfer in 60 seconds before speed, transfer in 60 seconds after number, transfer preceding speed, transfer in 60 seconds
Various transfer modes.
8. a kind of mode of transportation based on supporting vector machine model according to claim 7 changes to Activity recognition method,
Be characterized in that: the various transfer modes include walking transfer bicycle, bus is changed in walking, car is changed in walking, certainly
Driving transfer walking, bus transfering walking and car change to walking.
9. a kind of mode of transportation based on supporting vector machine model according to claim 1 changes to Activity recognition method,
Be characterized in that: the relevant information that behavior is changed to described in step 6 includes transfer time, place and number.
10. a kind of mode of transportation based on supporting vector machine model according to claim 1 changes to Activity recognition method,
Be characterized in that: the data base management system is the oracle database constructed using SaaS framework.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910583174.8A CN110276563A (en) | 2019-07-01 | 2019-07-01 | A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910583174.8A CN110276563A (en) | 2019-07-01 | 2019-07-01 | A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110276563A true CN110276563A (en) | 2019-09-24 |
Family
ID=67962760
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910583174.8A Pending CN110276563A (en) | 2019-07-01 | 2019-07-01 | A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110276563A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110675074A (en) * | 2019-09-27 | 2020-01-10 | 交通运输部公路科学研究所 | Travel target point identification method and device, and model development and evaluation method and device |
CN112185580A (en) * | 2020-08-20 | 2021-01-05 | 同济大学 | Travel risk information processing method based on cloud |
CN112542045A (en) * | 2020-12-01 | 2021-03-23 | 江苏欣网视讯软件技术有限公司 | Method and system for identifying traffic travel mode based on mobile phone signaling |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1870869A2 (en) * | 2006-06-19 | 2007-12-26 | Navteq North America, LLC | Traffic data collection with probe vehicles |
CN103310287A (en) * | 2013-07-02 | 2013-09-18 | 北京航空航天大学 | Rail transit passenger flow predicting method for predicting passenger travel probability and based on support vector machine (SVM) |
CN104537052A (en) * | 2014-12-26 | 2015-04-22 | 西南交通大学 | Transportation transfer station recognition method based on wavelet analysis modulus maximum algorithm |
-
2019
- 2019-07-01 CN CN201910583174.8A patent/CN110276563A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1870869A2 (en) * | 2006-06-19 | 2007-12-26 | Navteq North America, LLC | Traffic data collection with probe vehicles |
CN103310287A (en) * | 2013-07-02 | 2013-09-18 | 北京航空航天大学 | Rail transit passenger flow predicting method for predicting passenger travel probability and based on support vector machine (SVM) |
CN104537052A (en) * | 2014-12-26 | 2015-04-22 | 西南交通大学 | Transportation transfer station recognition method based on wavelet analysis modulus maximum algorithm |
Non-Patent Citations (1)
Title |
---|
周超然: "基于大规模GPS轨迹数据的活动链信息分析方法研究", 《中国博士学位论文全文数据库 工程科技II辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110675074A (en) * | 2019-09-27 | 2020-01-10 | 交通运输部公路科学研究所 | Travel target point identification method and device, and model development and evaluation method and device |
CN112185580A (en) * | 2020-08-20 | 2021-01-05 | 同济大学 | Travel risk information processing method based on cloud |
CN112185580B (en) * | 2020-08-20 | 2022-10-25 | 同济大学 | Travel risk information processing method based on cloud |
CN112542045A (en) * | 2020-12-01 | 2021-03-23 | 江苏欣网视讯软件技术有限公司 | Method and system for identifying traffic travel mode based on mobile phone signaling |
CN112542045B (en) * | 2020-12-01 | 2021-09-24 | 江苏欣网视讯软件技术有限公司 | Method and system for identifying traffic travel mode based on mobile phone signaling |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102607553B (en) | Travel track data-based stroke identification method | |
CN101350012B (en) | Method and system for matching address | |
CN110276563A (en) | A kind of mode of transportation transfer Activity recognition method based on supporting vector machine model | |
CN110414732B (en) | Travel future trajectory prediction method and device, storage medium and electronic equipment | |
CN103106280B (en) | A kind of range query method of uncertain space-time trajectory data under road network environment | |
CN102332210B (en) | Method for extracting real-time urban road traffic flow data based on mobile phone positioning data | |
CN102102992B (en) | Multistage network division-based preliminary screening method for matched roads and map matching system | |
CN107392245A (en) | A kind of taxi trajectory clustering algorithm Tr OPTICS | |
CN104159189A (en) | Resident trip information obtaining method based on intelligent mobile phone | |
CN105307121A (en) | Information processing method and device | |
CN111522892B (en) | Geographic element retrieval method and device | |
CN113360587B (en) | Land surveying and mapping equipment and method based on GIS technology | |
CN110727714A (en) | Resident travel feature extraction method integrating space-time clustering and support vector machine | |
CN110110243A (en) | A kind of historical track destination prediction technique based on echo state network | |
CN114265833A (en) | Vehicle track guiding geographic space-time characteristic visualization analysis platform | |
CN116437291A (en) | Cultural circle planning method and system based on mobile phone signaling | |
CN111242352A (en) | Parking aggregation effect prediction method based on vehicle track | |
CN115412857A (en) | Resident travel information prediction method | |
CN106980029B (en) | Vehicle overspeed judgment method and system | |
CN111405461A (en) | Wireless indoor positioning method for optimizing equal-interval fingerprint sampling number | |
CN113888867A (en) | Parking space recommendation method and system based on LSTM position prediction | |
CN112699201A (en) | Navigation data processing method and device, computer equipment and storage medium | |
CN105717517B (en) | A kind of vehicle-mounted big dipper multi-mode GNSS high accuracy roads basis collecting method | |
CN106991404B (en) | Ground mulching update method and system based on crowd-sourced geodata | |
CN113626729B (en) | Method and equipment for determining interest point information |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190924 |
|
RJ01 | Rejection of invention patent application after publication |