CN105869398A - Road traffic open degree judging method based on K-means cluster - Google Patents

Abstract

A road traffic open degree judging method based on a K-means cluster includes the steps of clearing away invalid data through a basic data threshold, so data distribution is more concentrated after data is primarily cleared away, further processing of data is easy, efficiency of data analysis is improved and result precision is enhanced; obtaining initial K values according to the highest limit speed of a road to correspond to different traffic states respectively; clustering different vehicle-mounted GPS data to different K value points through the K-means cluster, analyzing the number of data contained by each K value point, and making the maximum number serve as the index of a current road traffic state. By clustering and analyzing the vehicle-mounted GPS speed data, road network traffic state information is collected, implementation cost is low and technological difficulty is small.

Description

A kind of road traffic unimpeded degree determination methods based on K-means cluster

Technical field

The present invention relates to a kind of the coast is clear degree judge, computer application field, in particular, Yi Zhongji Road traffic unimpeded degree determination methods in K-means cluster.

Background technology

Recently as the quickening of Urbanization in China, the surge of vehicles number in city, urban environment shape Condition is the severeest, cannot alleviate current traffic pressure by urban road infrastructure construction merely.Intelligence is handed over It is connected with and helps ensure the unimpeded of intersection, reach to alleviate the purpose of traffic congestion, and then solve to a certain extent Determine the social problems such as urban traffic accident, traffic jam, environmental pollution and energy resource consumption, wherein road traffic shape The collection of state information is the basis realizing intelligent transportation.

City real vehicles driving road-condition is complicated, vehicle travel process can comprise acceleration, deceleration, at the uniform velocity, quiet Only four kinds of states, the data gathered are then the set of these three data, in some data in particular cases also Cannot function as judging when data exception that the fault of the foundation of road conditions, such as collecting device causes, personnel get off Stationary vehicle data (speed is zero), vehicle peccancy are driven over the speed limit (speed is much larger than speed limit), vehicle is promptly made Speed data (such as running into pedestrian to bring to a halt) during Dong, because the change of these in particular cases car speeds Change is not that the odjective cause of traffic causes, it is therefore necessary to some the most invalid data rejected.

Same road vehicle quantity is more, and the velocity information speed gathered differs, can not direct reaction Go out the traffic status of road, the vehicle speed information of same road can be carried out by K-means clustering algorithm Sort out, thus judge this road current state in certain time period.Represent that the description of traffic behavior generally includes Traffic density, passage rate, occupation rate etc., although these factors can reflect the current shape of a road State, but traffic density and occupation rate etc. are difficult to measure.

Summary of the invention

In order to overcome by the method such as traditional video monitoring obtain road traffic state information implementation cost higher and The deficiency that technical difficulty is bigger, invention herein provides a kind of unobstructed degree of road traffic based on K-means cluster to sentence Disconnected method, by carrying out cluster analysis to vehicle GPS speed data, it is achieved road grid traffic passes through status information Gathering, implementation cost is relatively low and technical difficulty is less.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of road traffic unimpeded degree determination methods based on K-means cluster, it is characterised in that: described detection Method comprises the following steps::

1) cleaning invalid data:

1.1) if car speed is that the data of zero are less than the 10% of total data on certain moment road, then it is assumed that vehicle Stopping is not that traffic congestion causes, it is judged that this data invalid；

1.2) if car speed is much larger than road speed limit, then it is assumed that this rule-breaking vehicle is driven, or collecting device Breaking down and cause gathering data exception, this type of data judging is invalid；

2) K value is selected:

Assuming that road exists N kind state, wherein the representation speed of every kind of state is V_n, by this N kind state Representation speed be inserted into initial data and concentrate that to carry out K-means cluster calculation as initial K value point quick Data set is gathered N number of apoplexy due to endogenous wind, then the n-th K value K_nChoose as follows；

K_n=V_n, n=1,2,3 N (1)

Wherein K_nValue along with the difference of road may be different, with the Maximum speed limit V of road_maxFor upper Limit, carries out speed limit decile calculating, and takes speed V of correspondence position_nAs K_nValue, speed V_nAccording to Equation below obtains:

$V_n=\frac{V_{max}\×(n-1)}{N},n=1,2,3...N---\left(2\right)$

3) cluster calculation:

3.1) N number of central point K is first selected₁,K₂,K₃···K_N, the most each K is worth exploitation such as formula (1) Shown in, they represent the corresponding current state of road respectively, and arranging iterations is g；

3.2) velocity information is traveled through successively, and the difference size of versus speed and central point, it is classified as difference minimum Apoplexy due to endogenous wind, it is judged that formula is as follows:

S_i-K_n=min{ | S_i-K₁|,|S_i-K₂|,|S_i-K₃|···|S_i-K_N|, n=1,2,3 N (3)

Wherein S_iRepresent i-th speed data, K_nTable the n-th K value, if meeting above-mentioned condition, then S is described_iBelong to the n-th class；

3.3) after iterative computation g time, adding up the sample size that each apoplexy due to endogenous wind comprises successively, the maximum class of quantity is wrapped

The K value point contained represents the current state of road.

The design thinking of the present invention is: same road vehicle quantity is more, and the velocity information speed gathered Differ, the traffic status of road can not be gone out by direct reaction, by K-means clustering algorithm can by with along with The vehicle speed information on road is sorted out, thus judges that the traffic in certain time period of this road is passed through state.

Beneficial effects of the present invention is mainly manifested in: carry out K-means cluster by gps data vehicle-mounted to road Analyze, vehicle speed information is converted into road conditions information more intuitively, solves road vehicle GPS number According to state dispersion can not the problem of direct reaction road conditions, the simplest K value selection has higher meter Calculate efficiency, the requirement to real-time in actual application can be met.

Accompanying drawing explanation

Fig. 1 is whole speed scatterplot.

Fig. 2 is the scatterplot after data scrubbing.

Fig. 3 is the flow chart of speed cluster.

Fig. 4 is cluster result schematic diagram.

Fig. 5 is the flow chart of a kind of road traffic unimpeded degree determination methods based on K-means cluster.

Detailed description of the invention

The present invention will be further described below in conjunction with the accompanying drawings.

With reference to Fig. 1～Fig. 5, a kind of road traffic unimpeded degree determination methods based on K-means cluster, including with Lower step:

1) cleaning invalid data:

Certain road is vehicle speed information data instance in 15 minutes, speed scatterplot distributed intelligence as it is shown in figure 1, Data that car speed on certain moment road is zero data less than the 10% of total data are rejected, by car Speed is rejected more than the data of road speed limit；

Result after cleaning as in figure 2 it is shown, wherein figure orbicular spot be invalid data, be placed in the figure lower left corner Because the data that speed is zero and speed is zero are judged as invalid data less than the 10% of total data, and It is placed in the speed in the figure upper right corner because being considered as hypervelocity or equipment event more than the speed limit of common road The invalid data that causes of barrier, after data are tentatively cleaned, the distribution of data is more concentrated, beneficially data enter one Step processes, and improves the efficiency of data analysis and strengthens result precision；

2) K value is selected:

Assuming that road exists N kind state, wherein the representation speed of every kind of state is V_n, the n-th K value Choose as follows:

K_n=V_n, n=1,2,3 N (1)

Maximum speed limit V with road_maxFor the upper limit, speed limit is carried out decile calculating, and takes the speed of correspondence position Degree V_nAs K_nValue, speed V_nCan obtain according to equation below:

$V_n=\frac{V_{max}\×(n-1)}{N},n=1,2,3...N---\left(2\right)$

3) cluster calculation:

Cluster calculation process is as shown in Figure 3；

3.1) N number of central point K is first selected₁,K₂,K₃···K_n, the most each K is worth exploitation such as formula (1) Shown in, they represent the corresponding current state of road respectively, and arranging iterations is g；

3.2) velocity information is traveled through successively, and the difference size of versus speed and central point, it is classified as difference minimum Apoplexy due to endogenous wind, it is judged that formula is as follows:

S_i-K_n=min{ | S_i-K₁|,|S_i-K₂|,|S_i-K₃|···|S_i-K_N|, n=1,2,3 N (3)

Wherein S_iRepresent i-th speed data, K_nTable the n-th K value, if meeting above-mentioned condition, then S is described_iBelong to the n-th class；

3.4) after iterative computation g time, adding up the sample size that each apoplexy due to endogenous wind comprises successively, the maximum class of quantity is wrapped The K value point contained represents the current state of road；

Cluster result is as shown in Figure 4；

56 vehicle GPS information in certain road 15min are carried out cluster analysis by this example, a kind of based on The road traffic unimpeded degree determination methods of K-means cluster, comprises the following steps:

1) cleaning invalid data:

The data that car speed on this road is zero are less than the 10% of total data by certain road speed limit 50km/h Data reject, by car speed more than road speed limit data reject, after cleaning, valid data are 44. 2) K value is selected:

Certain road speed limit is 50km/h, selects K₁,K₂,K₃,K₄Four values carry out cluster fortune as central point Calculate, represent respectively actual traffic road conditions be impassable, block up, normal, unobstructed four kinds of situations, pass through Judge four kinds of sizes clustered to judge the current state of road, four K values are the most as follows:

$\left\{\begin{array}{c}{K}_1=0km/h\\ K_2=\frac{50}{4}=12.5km/h\\ K_3=\frac{V_{max}}{2}=25km/h\\ K_4=\frac{3\×V_{\max }}{4}37.5km/h\end{array}\right.---\left(1\right)$

3) cluster calculation:

3.1) 4 central point K are first selected₁,K₂,K₃,K₄, the most each K is worth exploitation such as formula (1) Shown in, they represent the corresponding current state of road respectively, and taking iterations g is 100；

3.2) velocity information is traveled through successively, and the difference size of versus speed and central point, it is classified as difference minimum Apoplexy due to endogenous wind, it is judged that formula is as follows:

S_i-K_n=min{ | S_i-K₁|,|S_i-K₂|,|S_i-K₃|···|S_i-K_N|, n=1,2,3 N (2)

Wherein S_iRepresent i-th speed data, K_nTable the n-th K value, if the condition of meeting a cassation, then S is described_iBelong to the n-th class；

3.3) iteration 100 times, complete cluster calculation；

Cluster result data are as shown in table 1 below, the most altogether comprise vehicle speed information 44, Qi Zhongju Class is respectively 1,2,38 and 3 to the quantity of corresponding K value point, therefore can draw this road Road within this time period of 15 minutes the speed of most of vehicles all within normal range, therefore this road Road conditions are normal.

Table 1

The excellent effect of optimization that the embodiment that the present invention is given that described above is shows, it is clear that this Bright it is not only suitable for above-described embodiment, without departing from essence spirit of the present invention and without departing from involved by flesh and blood of the present invention And it can be done many variations on the premise of content and be carried out.

Claims (1)

1. a road traffic unimpeded degree determination methods based on K-means cluster, it is characterised in that: described detection Method comprises the following steps::

1) cleaning invalid data:

If on 1.1 certain moment road, car speed is that the data of zero are less than the 10% of total data, then it is assumed that vehicle Stopping is not that traffic congestion causes, it is judged that this data invalid；

If 1.2 car speeds are much larger than road speed limit, then it is assumed that this rule-breaking vehicle is driven, or collecting device Breaking down and cause gathering data exception, this type of data judging is invalid；

2) K value is selected:

Assuming that road exists N kind state, wherein the representation speed of every kind of state is V_n, by this N kind state Representation speed be inserted into initial data and concentrate that to carry out K-means cluster calculation as initial K value point quick Data set is gathered N number of apoplexy due to endogenous wind, then the n-th K value K_nChoose as follows；

K_n=V_n, n=1,2,3 N (1)

Wherein K_nValue along with the difference of road may be different, with the Maximum speed limit V of road_maxFor upper Limit, carries out speed limit decile calculating, and takes speed V of correspondence position_nAs K_nValue, speed V_nAccording to Equation below obtains:

$V_n=\frac{V_{max}\×(n-1)}{N},n=1,2,3...N---\left(2\right)$

3) cluster calculation:

3.1) N number of central point K is first selected₁,K₂,K₃···K_N, the most each K is worth exploitation such as formula (1) Shown in, they represent the corresponding current state of road respectively, and arranging iterations is g；

3.2) velocity information is traveled through successively, and the difference size of versus speed and central point, it is classified as difference minimum Apoplexy due to endogenous wind, it is judged that formula is as follows；

S_i-K_n=min{ | S_i-K₁|,|S_i-K₂|,|S_i-K₃|···|S_i-K_N|, n=1,2,3 N (3)

Wherein S_iRepresent i-th speed data, K_nTable the n-th K value, if meeting above-mentioned condition, then S is described_iBelong to the n-th class；

3.3) after iterative computation g time, adding up the sample size that each apoplexy due to endogenous wind comprises successively, the maximum class of quantity is wrapped The K value point contained represents the current state of road.