CN105389996A - Traffic operation condition characteristic parameter extraction method based on big data - Google Patents

Abstract

The invention discloses a traffic operation condition characteristic parameter extraction method based on big data and belongs to the technical field of traffic information. The traffic big data comprises mobile phone signaling data, GPS data and section traffic flow data based on videos and coils; and the traffic operation condition characteristic parameters comprise vehicle speed and road flow. The method comprises the following main steps: 1) extracting the traffic big data comprising the mobile phone signaling data, the GPS data and the section traffic flow data based on videos and coils; 2) matching the mobile phone data to a road network by adopting a mobile phone signaling data and GIS road network matching method based on Thiessen polygon; 3) calculating the vehicle speed by adopting the mobile phone data, and with the help of the GPS data, the precision of the computation speed is improved; and 4) establishing a relation model between the high-precision flow data obtained based on video data and coil data and the flow data extracted based on the mobile phone data, and extracting traffic flow information. The method can finish city large-scale traffic operation condition parameter acquisition within a short time.

Description

A kind of traffic circulation characteristic condition parameter extracting method based on large data

Technical field

The present invention relates to a kind of method that traffic circulation characteristic condition parameter extracts, especially a kind of traffic circulation characteristic condition parameter extracting method based on large data, belongs to transport information technical field.

Background technology

Traffic problems are products of Development of Urbanization.Along with the quickening of urbanization process, traffic problems are also on the rise, and wherein, traffic jam issue is one of the most serious problem, and people more and more pay close attention to how to administer traffic congestion.At present, by reasonable distribution and utilize existing road resource to alleviate the traffic congestion of part way, be a kind of effective means.Reasonable distribution and utilize the prerequisite of existing road resource to be exactly grasp to traffic circulation state.The method of current collection traffic circulation state mainly contains two kinds: fixed point gathers and flow acquisitions.Fixed point gathers and gathers mainly through monitors such as fixed position installation coil monitor, ultrasonic monitor, infrared observation device, microwave monitor, video monitor, and the data that this mode obtains are relatively accurate but acquisition range is little, and data volume is few.Flow acquisitions mainly refers to Floating Car collection, and the degree of accuracy that this mode obtains data places one's entire reliance upon the quantity of Floating Car on road network, and after Floating Car quantity is less than certain value, the precision obtaining data will reduce greatly.How to obtain comprehensively, in real time, accurately traffic circulation data become people pay close attention to problem.In recent years, along with the fast development of large data, study based on the acquisition technique of the traffic circulation data of large data, how to excavate mobile phone location data, GPS location data, analyze the traffic circulation characteristic condition parameters such as traffic circulation speed, the magnitude of traffic flow based on data such as video, the traffic flows of coil section and become a study hotspot.

At present, be also in the research primary stage by traffic circulation characteristic condition parameters such as mobile phone location data analysis traffic circulation speed, the magnitudes of traffic flow both at home and abroad, occurred part achievement in research.Speed calculates the real time signal data sent from all mobile phones in cell phone network acquisition time interval by Fixed Time Interval, obtains virtual-sensor section and travelling speed according to the real time signal data that i-th mobile phone sends; Average link speed calculates the speed calculation method having and adopt two-stage three map-matching methods and one to be called " LAC serial method ", also there is Negotiation speed gradual change segmentation to carry out difference processing to the travel speed of mobile phone terminal on road simultaneously, and utilize the distribution of the travelling speed interval probability of history mobile phone speed sample to be weighted average computation to section travelling speed and hourage.

And for the magnitude of traffic flow, the telecommunication flow information, traffic noise data, road type information and the acquisition time type information that are obtained by data input step that have set up artificial nerve network model, and then obtain traffic flow information; Also there is the related coefficient by calculating data in mobile phone and road traffic flow data, utilizing data in mobile phone and the related coefficient of Real-Time Monitoring, the traffic flow data of prediction current slot.

There is following shortcoming in the method being obtained the traffic circulation state parameters such as traffic circulation speed, the magnitude of traffic flow above by data in mobile phone:

1) mobile phone location Data Matching is not high to the precision of road net, to such an extent as to loss of data or path adaptation mistake;

2) mobile phone carrier wide material sources, both can from the other crowd on foot of road, also the passenger of bus and car can be carried out, also can from the driver driving bicycle, but only have carrier to be that cellphone subscriber on road in motor vehicle is only the sample that can be used for road conditions and calculate, because carrier is difficult to distinguish, cause the traffic that the traffic information calculated can not reflect road faithfully;

3) do not consider the difference of road track classification when calculating road average velocity, by all sections speed, average or weighted mean all really can not reflect the real average velocity of road.

4) calculating of road Traffic Volume is more coarse, for the screening of data in mobile phone and the even long-term vehicle flow monitoring of short-term perfect not.

Summary of the invention

For the defect that above-mentioned prior art exists, the invention provides a kind of traffic circulation status monitoring parameter extracting method based on large data, the large data of the traffic adopted mainly comprise mobile phone location data, gps data, based on video, coil section traffic flow data, the traffic circulation state parameter of monitoring comprises speed and flow.Mainly comprise following step: the 1) extraction of the large data of traffic, its comprise mobile phone location data, gps data, based on video, coil section traffic flow data; 2) data in mobile phone and road network, takes to match on road network based on the mobile phone location data of Thiessen polygon and GIS road network method by data in mobile phone; 3) extraction rate information take data in mobile phone as core, and other data are auxiliary, improves mobile phone velocity accuracy, optimizes the VELOCITY EXTRACTION model based on data in mobile phone and VELOCITY EXTRACTION result; 4) extract flow information, set up the relational model that the high-precision flow data and the data in mobile phone that obtain based on video data, loop data extract data on flows, optimize the magnitude of traffic flow extraction model based on data in mobile phone and magnitude of traffic flow extraction result.

Concrete technical scheme of the present invention is as follows:

Step 1, the extraction of the large data of traffic

The large data of traffic of the present invention comprise mobile phone location data, gps data, based on video, coil section traffic flow data.Extract these data specific practices as follows:

Step 1.1, is divided into multiple section by road, and records the information such as section numbering, position coordinates, milimeter number in each section, collects and covers each section associated base stations number, record base station numbering and position coordinates;

Step 1.2, selectively chooses the data in mobile phone with traffic counts associated base stations in cycle t from database, and the data in mobile phone of acquisition has Customs Assigned Number (IMSI), communication type, base station numbering, switching time;

Step 1.3, chooses gps data relevant to traffic counts in cycle t from database;

Step 1.4, chooses in cycle t relevant to traffic counts based on video, coil section traffic flow data from database.

Step 2, data in mobile phone and road network

Take the mobile phone location data based on Thiessen polygon and GIS road network method, specific practice is as follows:

Step 2.1, all for covering traffic counts adjacent base stations are linked to be triangle, make the perpendicular bisector on each limit of these triangles, so the some perpendicular bisectors around each base station just surround a polygon, obtain the lone base station comprised in this polygon.This polygon is called Thiessen polygon, and this base station is called Tyson base station;

Step 2.2, carries out a classification of the screening of IMSI number of iterations and Event_ID by the data in mobile phone in Tyson base station, mobile phone signaling data after acquisition relevant treatment.

Step 2.3, carries out GIS map coordinate matching by the Tyson base station obtained in step 2.1 and step 2.2 and signaling data, obtains Tyson signaling data;

Step 2.4, selects the Tyson signaling data of every IMSI, according to time sequence.As:

IMSI＝xxxxxxxx

Time1，CI1，location1，length1

Time2，CI2，location2，length2

Time3，CI3，location3，length3

…

Mobile phone location data must be arrived.

Step 3, extraction rate information

In extraction rate process, fully demonstrate the calculating of screening process to speed data and the most rational Road average-speed, specific practice is as follows:

Step 3.1, processes the mobile phone location data that step 2.1 obtains.Select the time location data of position i and the time location data of a rear position i+1, calculate mistiming (DeltaT between the two _i) and range difference (DeltaD _i), speed V _i=DeltaD _i/ DeltaT _i;

Step 3.2, the section all position and speeds being matched starting point section, terminal section and live through;

Step 3.3, through the data such as high-precision GPS and mobile phone speed data convergence analysis and selective retention distance at X apart from above section speed data and selective retention, signaling data is separated by section speed data more than the X time, obtains constant value speed data;

Step 3.4, carries out GMM statistics to constant value speed data.If have speedway and low speed carriage way in signaling area, added up by GMM and distinguish speedway average velocity and low speed carriage way average velocity, specific practice is as follows:

Step 3.4.1, calculates the weight α in the average μ of constant value speed on each bar section on road, standard deviation sigma and each bar section;

Step 3.4.2, adds up every bar section speed V _i∈ (μ-σ, μ+σ), demarcating these speed is vehicle velocity V on speedway on this section _k, road speedway average velocity is calculated as:

Step 3.4.3, adds up every bar section speed demarcating these speed is vehicle velocity V on low speed carriage way on this section _d, road speedway average velocity is calculated as:

Step 4: extract flow information

Set up the high-precision flow data that obtain based on video data, loop data and data in mobile phone extracts the relational model of data on flows, thus better obtain traffic flow information, specific practice is as follows:

Step 4.1, to the mobile phone location data (N that step 2.1 obtains _c) process, select the mobile phone location data (N that there is the fixed point such as video monitor or coil monitor collector section _s);

Step 4.2, obtains the video data of same period t, loop data etc. from database, obtains the magnitude of traffic flow (N of the higher road sections part position of precision _z);

Step 4.3, the coefficient of relationship (r) between the magnitude of traffic flow of collector of fixing a point in calculating section and the data in mobile phone of same position, $r=\frac{\underset{i=1}{\overset{n}{\Σ}}(N_s^{\left(i\right)}-\stackrel{\‾}{N_s})(N_z^{\left(i\right)}-\stackrel{\‾}{N_z})}{\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(N_s^{\left(i\right)}-\stackrel{\‾}{N_s})}^2*\underset{i=1}{\overset{n}{\Σ}}{(N_z^{\left(i\right)}-\stackrel{\‾}{N_z})}^2}},$ Both r just illustrates close to 1, correlativity is strong, performs step 4.4, if correlativity is poor, performs step 4.6;

Step 4.4, calculates the high-precision flow data that obtain based on video data, loop data and data in mobile phone extracts the relational model of data on flows, N=a+bN _c+ cN _c ², wherein, a, b, c meet following equation:

$\left[\begin{array}{c}n,\underset{i=1}{\overset{n}{\Σ}}{N}_s,\underset{i=1}{\overset{n}{\Σ}}{N}_s^2\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s,\underset{i=1}{\overset{n}{\Σ}}{N}_s^2,\underset{i=1}{\overset{n}{\Σ}}{N}_s^3\\ \underset{j=1}{\overset{n}{\Σ}}{N}_s^2,\underset{i=1}{\overset{n}{\Σ}}{N}_s^3,\underset{i=1}{\overset{n}{\Σ}}{N}_s^4\end{array}\right]\·\left[\begin{array}{c}a\\ b\\ c\end{array}\right]=\left[\begin{array}{c}\underset{i=1}{\overset{n}{\Σ}}{N}_s{N}_z\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^2{N}_z\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^3{N}_z\end{array}\right]$

Step 4.5, brings relational model into and just can obtain each section Real-Time Traffic Volume (N), N=a+bN by the data in mobile phone of each for section position each period _c+ cN _c ².

Step 4.6, the gap of data in mobile phone and actual traffic amount, without presenting irregularities, cannot well adopt data in mobile phone to calculate the magnitude of traffic flow, can only take the data determination magnitude of traffic flow that the fixed point such as video monitor or coil monitor collector obtains.

Accompanying drawing explanation

Fig. 1 is traffic circulation characteristic condition parameter extracting method general flow chart of the present invention

Fig. 2 is mobile phone signaling data of the present invention and road network algorithm flow chart

Fig. 3 is vehicle speed information extraction algorithm process flow diagram of the present invention

Fig. 4 is vehicle flow information extraction algorithm process flow diagram of the present invention

Specific embodiments

Be described in further detail to feature of the present invention and other correlated characteristic below in conjunction with accompanying drawing.

Step 1, the extraction of the large data of traffic

The large data of traffic of the present invention comprise mobile phone location data, gps data, based on video, coil section traffic flow data.Extract these data specific practices as follows:

Step 1.1, is divided into multiple section by road, and records the information such as section numbering, position coordinates, milimeter number in each section, collects and covers each section associated base stations number, record base station numbering and position coordinates;

Step 1.2, selectively chooses the data in mobile phone with traffic counts associated base stations in cycle t from database, and the data in mobile phone of acquisition has Customs Assigned Number (IMSI), communication type, base station numbering, switching time;

Step 1.3, chooses gps data relevant to traffic counts in cycle t from database;

Step 1.4, chooses in cycle t relevant to traffic counts based on video, coil section traffic flow data from database.

Step 2, data in mobile phone and road network

Take the mobile phone location data based on Thiessen polygon and GIS road network method, as shown in Figure 2, specific practice is as follows for algorithm flow:

Step 2.1, all for covering traffic counts adjacent base stations are linked to be triangle, make the perpendicular bisector on each limit of these triangles, so the some perpendicular bisectors around each base station just surround a polygon, obtain the lone base station comprised in this polygon.This polygon is called Thiessen polygon, and this base station is called Tyson base station;

Step 2.2, carries out a classification of the screening of IMSI number of iterations and Event_ID by the data in mobile phone in Tyson base station, mobile phone signaling data after acquisition relevant treatment.

Step 2.3, carries out GIS map coordinate matching by the Tyson base station obtained in step 2.1 and step 2.2 and signaling data, obtains Tyson signaling data;

Step 2.4, selects the Tyson signaling data of every IMSI, according to time sequence.As:

IMSI＝xxxxxxxx

Time1，CI1，location1，length1

Time2，CI2，location2，length2

Time3，CI3，location3，length3

…

Mobile phone location data must be arrived.

Step 3, extraction rate information

In extraction rate process, fully demonstrate the calculating of screening process to speed data and the most rational Road average-speed, as shown in Figure 3, specific practice is as follows for algorithm flow:

Step 3.1, processes the mobile phone location data that step 2.1 obtains.Select the time location data of position i and the time location data of a rear position i+1, calculate mistiming (DeltaT between the two _i) and range difference (DeltaD _i), speed V _i=DeltaD _i/ DeltaT _i;

Step 3.2, the section all position and speeds being matched starting point section, terminal section and live through;

Step 3.3, through the data such as high-precision GPS and mobile phone speed data convergence analysis and selective retention distance at X apart from above section speed data and selective retention, signaling data is separated by section speed data more than the X time, obtains constant value speed data;

Step 3.4, carries out GMM statistics to constant value speed data.If have speedway and low speed carriage way in signaling area, added up by GMM and distinguish speedway average velocity and low speed carriage way average velocity, specific practice is as follows:

Step 3.4.1, calculates the weight α in the average μ of constant value speed on each bar section on road, standard deviation sigma and each bar section;

Step 3.4.2, adds up every bar section speed V _i∈ (μ-σ, μ+σ), demarcating these speed is vehicle velocity V on speedway on this section _k, road speedway average velocity is calculated as:

Step 3.4.3, adds up every bar section speed demarcating these speed is vehicle velocity V on low speed carriage way on this section _d, road speedway average velocity is calculated as:

Step 4: extract flow information

Set up the high-precision flow data that obtain based on video data, loop data and data in mobile phone extracts the relational model of data on flows, thus better obtain traffic flow information, as shown in Figure 4, specific practice is as follows for specific algorithm flow process:

Step 4.1, to the mobile phone location data (N that step 2.1 obtains _c) process, select the mobile phone location data (N that there is the fixed point such as video monitor or coil monitor collector section _s);

Step 4.2, obtains the video data of same period t, loop data etc. from database, obtains the magnitude of traffic flow (N of the higher road sections part position of precision _z);

Step 4.3, the coefficient of relationship (r) between the magnitude of traffic flow of collector of fixing a point in calculating section and the data in mobile phone of same position, $r=\frac{\underset{i=1}{\overset{n}{\Σ}}(N_s^{\left(i\right)}-\stackrel{\‾}{N_s})(N_z^{\left(i\right)}-\stackrel{\‾}{N_z})}{\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(N_s^{\left(i\right)}-\stackrel{\‾}{N_s})}^2*\underset{i=1}{\overset{n}{\Σ}}{(N_z^{\left(i\right)}-\stackrel{\‾}{N_z})}^2}},$ Both r just illustrates close to 1, correlativity is strong, performs step 4.4, if correlativity is poor, performs step 4.6;

Step 4.4, calculates the high-precision flow data that obtain based on video data, loop data and data in mobile phone extracts the relational model of data on flows, N=a+bN _c+ cN _c ², wherein, a, b, c meet following equation:

$\left[\begin{array}{c}n,\underset{i=1}{\overset{n}{\Σ}}{N}_s,\underset{i=1}{\overset{n}{\Σ}}{N}_s^2\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s,\underset{i=1}{\overset{n}{\Σ}}{N}_s^2,\underset{i=1}{\overset{n}{\Σ}}{N}_s^3\\ \underset{j=1}{\overset{n}{\Σ}}{N}_s^2,\underset{i=1}{\overset{n}{\Σ}}{N}_s^3,\underset{i=1}{\overset{n}{\Σ}}{N}_s^4\end{array}\right]\·\left[\begin{array}{c}a\\ b\\ c\end{array}\right]=\left[\begin{array}{c}\underset{i=1}{\overset{n}{\Σ}}{N}_s{N}_z\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^2{N}_z\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^3{N}_z\end{array}\right]$

Step 4.5, brings relational model into and just can obtain each section Real-Time Traffic Volume (N), N=a+bN by the data in mobile phone of each for section position each period _c+ cN _c ².

Step 4.6, the gap of data in mobile phone and actual traffic amount, without presenting irregularities, cannot well adopt data in mobile phone to calculate the magnitude of traffic flow, can only take the data determination magnitude of traffic flow that the fixed point such as video monitor or coil monitor collector obtains.

Claims (6)

1., based on a traffic circulation characteristic condition parameter extracting method for large data, it is characterized in that, comprise the following steps:

Step 1, extracts and comprises mobile phone signaling data, gps data, large data of traffic based on video, coil section traffic flow data;

Step 2, takes to match on road network based on the mobile phone signaling data of Thiessen polygon and GIS road network method by data in mobile phone;

Step 3, takes data in mobile phone to calculate car speed, and is auxiliary with gps data, improve the precision of computing velocity, and considers that road track speed limit is different, takes divided lane to divide speed;

Step 4, sets up the relational model that the high-precision flow data and the data in mobile phone that obtain based on video data, loop data extract data on flows, extraction flow information.

2. the traffic circulation characteristic condition parameter extracting method based on large data according to claim 1, it is characterized in that, first the large data of described traffic are extract mobile phone signaling data, city large-scale traffic circulation state parameter collection can be completed at short notice based on mobile phone signaling data, next extracts gps data, dynamic data static in interior existing Urban Transportation based on video, coil section traffic flow data, improves traffic circulation state parameter acquisition precision based on these data.

3. the traffic circulation characteristic condition parameter extracting method based on large data according to claim 1, is characterized in that, described based on the mobile phone signaling data of Thiessen polygon and the method for GIS road network is:

Step 2.1, all for covering traffic counts adjacent base stations are linked to be triangle, make the perpendicular bisector on each limit of these triangles, so the some perpendicular bisectors around each base station just surround a polygon, obtain the lone base station comprised in this polygon.This polygon is called Thiessen polygon, and this base station is called Tyson base station;

Step 2.2, carries out a classification of the screening of IMSI number of iterations and Event_ID by the data in mobile phone in Tyson base station, mobile phone signaling data after acquisition relevant treatment.

Step 2.3, carries out GIS map coordinate matching by the Tyson base station obtained in step 2.1 and step 2.2 and signaling data, obtains Tyson signaling data;

Step 2.4, selects the Tyson signaling data of every IMSI, according to time sequence.As:

IMSI＝xxxxxxxx

Time1，CI1，location1，length1

Time2，CI2，location2，length2

Time3，CI3，location3，length3

…

Mobile phone location data must be arrived.

4. the traffic circulation characteristic condition parameter extracting method based on large data according to claim 1, is characterized in that, the described car speed extracting method based on large data is:

Step 3.1, processes the mobile phone location data that step 2.1 obtains.Select the time location data of position i and the time location data of a rear position i+1, calculate mistiming (DeltaT between the two _i) and range difference (DeltaD _i), speed V _i=DeltaD _i/ DeltaT _i;

Step 3.2, the section all position and speeds being matched starting point section, terminal section and live through;

Step 3.3, through the data such as high-precision GPS and mobile phone speed data convergence analysis and selective retention distance at X apart from above section speed data and selective retention, signaling data is separated by section speed data more than the X time, obtains constant value speed data;

Step 3.4, carries out Gaussian statistics by the constant value speed of acquisition, considers that the speed limit in the different track of segment path is different, takes divided lane to divide speed.

5. the car speed extracting method based on large data according to claim 4, is characterized in that, the Gaussian statistics method that described car speed divided lane divides speed is:

Step 3.4.1, calculates the weight α in the average μ of constant value speed on each bar section on road, standard deviation sigma and each bar section;

Step 3.4.2, adds up every bar section speed V _i∈ (μ-σ, μ+σ), demarcating these speed is vehicle velocity V on speedway on this section _k, road speedway average velocity is calculated as:

Step 3.4.3, adds up every bar section speed demarcating these speed is vehicle velocity V on low speed carriage way on this section _d, road speedway average velocity is calculated as:

6. the traffic circulation characteristic condition parameter extracting method based on large data according to claim 1, is characterized in that, the described vehicle flow extracting method based on large data is:

Step 4.1, to the mobile phone location data (N that step 2.1 obtains _c) process, select the mobile phone location data (N that there is the fixed point such as video monitor or coil monitor collector section _s);

Step 4.2, obtains the video data of same period t, loop data etc. from database, obtains the magnitude of traffic flow (N of the higher road sections part position of precision _z);

Step 4.3, the coefficient of relationship (r) between the magnitude of traffic flow of collector of fixing a point in calculating section and the data in mobile phone of same position, both r just illustrates close to 1, correlativity is strong, performs step 4.4, if correlativity is poor, performs step 4.6;

Step 4.4, calculates the high-precision flow data that obtain based on video data, loop data and data in mobile phone extracts the relational model of data on flows, N=a+bN _c+ cN _c ², wherein, a, b, c meet following equation:

$\left[\begin{array}{c}n,\underset{i=1}{\overset{n}{\Σ}}{N}_s,\underset{i=1}{\overset{n}{\Σ}}{N}_s^2\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s,\underset{i=1}{\overset{n}{\Σ}}{N}_s^2,\underset{i=1}{\overset{n}{\Σ}}{N}_s^3\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^2,\underset{i=1}{\overset{n}{\Σ}}{N}_s^3,\underset{i=1}{\overset{n}{\Σ}}{N}_s^4\end{array}\right]\·\left[\begin{array}{c}a\\ b\\ c\end{array}\right]=\left[\begin{array}{c}\underset{i=1}{\overset{n}{\Σ}}{N}_s{N}_z\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^2{N}_z\\ \underset{i=1}{\overset{n}{\Σ}}{N}_s^3{N}_z\end{array}\right]$

Step 4.5, brings relational model into and just can obtain each section Real-Time Traffic Volume (N), N=a+bN by the data in mobile phone of each for section position each period _c+ cN _c ².

Step 4.6, the gap of data in mobile phone and actual traffic amount, without presenting irregularities, cannot well adopt data in mobile phone to calculate the magnitude of traffic flow, can only take the data determination magnitude of traffic flow that the fixed point such as video monitor or coil monitor collector obtains.